Introduction: David Bowie and the Art of Failure
1/10/2016: David Bowie is suddenly gone, using Death itself as a final performance. And we fell into mourning: my girlfriend Linda had introduced me to all things Bowie, and now his absence felt like a hole in the world.
Bowie might have said "fill it with something weird", so I built this for Linda: A David Bowie Memorial End-grain Cutting Board in Matching Display Case. The idea fell out of some sad timing: Bowie died just before Linda's birthday, and I was already deep in the design of her original gift: a Union Jack cutting board inspired by Youtuber MTMWood. With Bowie's death I was reminded of his Earthling album cover and its iconic Union Jack greatcoat by Alexander McQueen (another loss! r.i.p. 2010). Could I merge the two? As I pushed the CAD around to match the coat, I gained new insight into Bowie's and McQueen's work, the politics above all: McQueen's Union Jack splits the standard down the middle and reverses it. The top and bottom are asymmetric, and the orientation 90 degrees off the flag's. This aggressive dismantling of a British icon, slashed with rips and tears--I never grasped those implications before digging in.
I should mention now, Dear Reader, that this Instructable veers from the typical step-by-step template. This entire project spanned ~4 mos, ~50 Jira tickets, 100s of pictures, and both 2D and 3D design assets in Sketchup + Illustrator. Before tackling that (this being my first Instructable), I'm pushing another angle: lessons in the Art of Failure. For that let me provide the briefest description of the board & case, then on to The Failures!
(that said, these steps & pix are in chronological order so if you've a jones for your own, they'll get you miles further than where I started).
Design and Outcome
The original photography for Earthling was actually shot on a white background (the green field of the album cover was added in post). My case is modeled on the original photo, with a white top that seems at first a simple recreation. Look closer, and you notice that the white, blue, and red of Bowie's coat are cut away to reveal matching wood hues beneath. This in fact was the central challenge: a cutting board precise enough to line up photographically across its span. The result is a surprising visual effect each time the door is lifted.
This level of registration required every piece of wood to be machined within +/- 0.005". For a woodworking noob like myself, that tiniest-of-margins brought a plague of problems (read on).
TL;DR : Study the artifacts of failure with a child's eye (no judgement). Iterate fast by failing cheaply, quickly, at very small scales. Iterate often to thicken your "failure skin" letting you push through anything. Work on something you love, for someone you love.
Cutting Board Materials:
- Maple (blue and white)
- Padauk (red)
- Ebony (black)
- Analine dye ("purplish-blue")
- White acrylic with clear acrylic windows and hinges (laser cut)
- Photo paper (laser cut)
- Unibody case, CNCed from 6 layers of 1/4" fin-ply
- Vacuum-formed, fused-felt tub: PETG sheet, polyester felt, closed cell foam
- 3D-printed positioning clips, mating to cutting board insets
- Torsion springs
- 1/16" stainless welding rod for door hinges
Step 1: Prime Cuts (Accuracy of Saws)
Band saw (worst):
Drawn by the siren song of a 1/32" kerf, the band saw promised thin cuts and less waste. And look! A fence! Fences=straight lines, yes? In fact, any fence on a band saw guarantees only the starting offset of a cut, beyond which the flexible blade can meander through long cuts.
Chop saw (better):
The rigid blade of a chop saw adds precision, but its various degrees of freedom (rotating base, lowering pivot, radial track) steal imprecision back.
Radial saw (even better):
Losing the chop saw's pivot drops a degree of freedom and gains some precision+repeatability.
Table saw (best):
One degree of freedom. A beefy fence and a sharp, thin-kerf blade was ultimately the best combo for max precision & min waste.
CNC (best++, where appropriate):
For consistent planing/surfacing down to 0.001", nothing beats CNC e.g. Shopbot or Shapeoko.
Lessons: to register so many layers, nearly everything in this project needed a CAM approach: Shopbot, 3D-printing, waterjet or laser cutter. For the cutting board, my workflow settled on the Shopbot and table saw: the saw is fast and gets you to +/-0.1", then Shopbot surfacing by 0.01" or 0.001".
Step 2: Under Pressure (Glue-ups)
With the cutting board's end-grain design, its pattern needed a total of 17 glue-up operations. Early attempts were mess: drips littering vice and table, and even worse: individual boards wandering from their starting positions like Billy Bob teeth!
Lessons: Wet glue-ups are fluid structures that will drift along unconstrained dimensions, so "3D clamping" with cauls is a must. Use catch paper to keep things neat. A good prep job is complex and rather pretty; the output is clean and straight; the tools remain clean.
Step 3: Cracked Actor (Planer Trouble)
New glue-ups are gnarly with drips. My cleanup approach was a classic S4S workflow (squared 4 sides) through the jointer & planer. Since it worked for the maple blocks, I proceeded confidently with my first, smaller block of paduak. As it disappeared into the planer's maw, the machine sputtered and banged as the block imploded under the feed roller. Holding the red shards afterwards, I was overcome with shame for ruining such a fine specimen. This was a living beautiful thing, beautiful still as lumber. And I killed it.
Lessons: only plane pieces long enough to span the rollers, and/or attach a sacrificial perimeter of scrap when in doubt. Such a perimeter is the ultimate solution to both roller pressure and tear-out.
Step 4: Chameleon (Wood Strip Sizing)
For some reason, it didn't occur to me to CNC the wood strips to thickness. But every other cockamamy idea occurred and I rat-holed on every one.
Early results with a belt sander and block were primitive: the slow abrasion I'd expected is really quite rapid at small scales of 0.01". Also lumpy over long spans.
I then spent several days building a screw-adjusted hand-sanding jig, the height of which controlled the workpiece's stick-out from a steel sanding plate: sand the wood flush, and the thickness is set! But ...
- I hadn't anticipated the deep trough curling the wood, so I had to also waterjet a 1/4" Al back-brace. Now the jig sorta worked, but ...
- The workpiece never held rigidly & sanding vibrations queered the height setting. Then it hit me ...
- "Am I sanding metal shavings into wood bound for the SawStop?" Not cool.
Sanding even that first piece, I realized "CNC sanding" was the answer. Even so, the jig's an awesome relic!
Ultimately I landed on the following workflow to CNC each strip to thickness:
- Face the strip sawed-side down on a pre-surfaced spoil board.
- Mount the strip with turner's tape (Spectape is my far-and-away favorite). Given the shallow cuts, milling forces are minimal so no need to overdo it with the tape.
- Run a 0.01" surfacing pattern with a 1/2" end mill.
- Repeat #3, adjusting start-Z as finely as 0.001" (e.g. ebony).
Lessons: the sanding jig was a functional failure but a CAM success! It married a water-jetted Aluminum base, a CNC'ed wood block, and a water-jetted steel top plate all from the same source geometry! And it bolted up perfect the first time which was pretty cool.
Step 5: Little Wonder (CNC Near Miss)
At the center of the cutting board are four angled blocks. The top and bottom pairs--mirror images of each other--were cut at 180 degree offsets from the same plank. Given their anchor position, size and squareness were paramount so CNC was the only option I considered to cut them out.
Thinking ahead, I'd left plenty of machining material in the parent board ... so I thought! While homing the first block's cutout run, I left a comfortable margin to the spoil material. I didn't take into account the added 1" of perimeter needed to clear a 1/2" end mill, and ended up squeezing the second block to the extreme opposite edge, clearing by 0.1".
Lessons:Never take slack for granted. Use it intentionally & bank the rest.
Step 6: Holy Holy (CNC Miss)
As time and value accumulate, so too the potential for catastrophe from even simple mistakes! For an instant, with the late-stage board mounted in the Shopbot for a final surfacing pass, I thought I'd blown it: instead of the 0.01" shave I'd programmed, the bit dove down 1/2" and started eating a trough up the middle. Remember direct hits in dodge-ball? I squealed just like that all the way to the emergency stop, then stared agape: did I just lose a month's work?
Thankfully I caught a moment of grace. Any edge-ward movement of the end-mill at that depth would have cracked the tape with explosive results, but I managed to abort before any turns (comparing cut length to feed rate, my reaction time was ~1.5s. Had I covered the e-stop per SOP: <0.5s).
Lucky also that the surfacing pattern spiraled out from the center, sparing the outer visible edges. Ultimately the damage submerged under an adjacent strip of padauk, a secret now only x-rays could tell (anyone in the SF Bay Area with one to borrow? I'm keen for a shot).
Lessons: Always cover the e-stop! I lapsed, given the repetition of many surfacing passes, and mistyped the next Z home.
Step 7: Time (Dying the Final Board)
I found a "purplish blue" aniline dye to match Bowie's jacket at Woodworker's Supply. I applied it liberally to the maple before glue-up, assuming it would penetrate to survive later finishing passes. Nope! ... stripped by the first pass through the jointer! I now faced the worrying prospect of dye-bleed on the final board.
So I set the board aside and dug into my pile of maple scraps. I set about getting a razor-sharp, bleed-free line which eventually came down to surface porosity and brush wetness. To achieve a sharp edge I first finish-sanded to 1200 grit to reduce porosity. This improved the masking tape seal dramatically, the final trick being a dry-as-possible brush: after each dip, I'd blot on paper bags until the brush was nearly dry. Through this buildup of light & fast-drying layers, the blue came in slowly while perfectly isolated.
Lessons: Take the time to feel comfortable. When uncomfortable, seek an approach that's tiny, incremental, and low-risk.
Step 8: Labyrinth (Case Total Fail)
The case gets its strength from an internal wood frame. My original design consisted of 54 laser-cut pieces, assembled into 9 frame members stacking 6 identical pieces each. These 9 members would assemble into the final frame.
The many-piece design was a no-starter. To begin with, the edges were charred from the laser meaning lots of sanding before dying could happen. Even so, I proceeded to assemble the 9 frame members, but the accumulation of errors soon discouraged any attempt at final assembly. Plus the abundance of seams foretold a wobbly mess.
Losing seams was key. As my thinking turned in that direction, I thought of automobile frames and the industry's move to unibody construction. Suddenly it was obvious: instead of many small pieces, I could simply pre-glue a full-size blank of solid wood, and CNC a seamless "unibody" frame of my own.
The result? 54 pieces to one, 14 seams to zero.
Lessons: K.I.S.S. (keep it seamless, stupid)
Step 9: Space Oddity (Another CNC Miss)
My first attempt to CNC the unibody frame was mis-homed, so the cut ran displaced for some seconds. I aborted, but the damage was difficult to assess: the run needed to remove the majority of the interior anyway, so I was maybe OK? I re-homed to "have a go" as the Ozzies say. The end result had a small hole, deemed patchable given its position in the largely-unseen rear wall ... wabi-sabi!
Lessons: I always precede real CNC runs with ghost runs to confirm my g-code. Unfortunately, here I aborted the ghost run before it completed its full perimeter (watching the time in a short Shopbot session). In reality I had only tested the Z-home (which was OK), not the XY-home (which was off). Now I always run at least one full Z-layer of g-code during my ghost run.
Step 10: Always Crashing in the Same Car (Four Clip Designs)
The case's felt liner is slightly larger than the cutting board it holds, leaving some slop. That's a problem since the see-through effect only works when the cutouts and cutting board are aligned! To fix the alignment, clips are installed which engage cutouts on each end of the cutting board.
The clips are 3D-printed and designed to match the cutouts. They're also spring loaded to "clip in" when the cutting board is lowered on them, with an opposing lever to "clip out" and release the board from the case.
The 3D model for these clips went through 4 complete redesigns, "complete" meaning 80%+ new geometry.
- Mechanical details, like pivot points, were too small for the resolution of my 3D printer. Total fail.
- A scaled-up version printed OK, but still too weak. For instance it incorporated a plastic spring which cracked more than flexed. Total fail.
- Wacky cam-shaped release levers. A fun modeling exercise, but also too small and weak. Total fail.
- v4 incorporated a steel spring, beefy components, and a 4-piece design to facilitate assembling the spring into the mechanism. Works OK.
The 4th clip iteration was built into the case, and works OK. However, it isn't strong enough to pass the "upside down shake" test. So of course I'm thinking of a v5 that ditches the spring for carbon fiber construction, with a cam lock to really grip the cutting board.
Lessons: I learned tons about 3D printing--specifically CAD for 3D printing--while iterating these clips. v1 showed me the effective resolution of 3D printing, namely that features <1mm aren't feasible despite looking feasible in CAD. At that scale, filament oozing introduces large std deviations so things don't fit together. At scales >10mm however, accuracy comes from the repeatability of stepper motors which is dead accurate. This was also my first mechanical device with moving parts, so I had to learn movement simulation in Sketchup (hard, keyframe plugins required). So hard, in fact, that I've since moved to Fusion 360 given its native support for Joints and Assemblies.
I should emphasize how gutted I was at the poor v1 attempt. It was useless, and the path to something useful seemed hopeless. In these moments I found it useful to write things down--as much for therapy as design review--and before long I'd find a toe-hold for the next try. Trust that you're learning (it may not feel like it), and improving (it may not look like it).
Step 11: Low (Vacuum Forming)
I had zero experience with vacuum forming before this project. I only knew that I needed a felt-lined, custom-fit lining and VF seemed the trick. Early attempts weren't to build a tub, rather just to learn Techshop's VF machine and process, so bad stuff happened like setting the temperature too high causing the PETG to bubble up towards the ceramic heaters. Or jumping the gun before the plastic had fully melted + sagged, which killed the fit.
My first attempt at an actual tub was ugly, but educational. I used the cutting board itself as a mold, with 3D-printed spacers stuck on to create bulges matching the case. Here's where I learned that the melting point of PLA is below that of PETG, so the spacers melted and collapsed. Also, the felt "jacket" I'd taped on showed large seams and overlaps. Lastly, with no vents in the mold, air pockets couldn't escape leading to random bubbles and folds. Still, the PETG+felt fused together as hoped.
Back to the Shopbot: my next attempt went "all in" with a fully CNCed mold of wood. The new mold was riddled with vent holes to shed air, and a new, hand-sewn felt jacket was closely fitted. Finally, a 3D-printed "skirt" was attached to flare the top edge. After these changes, the 2nd iteration was perfect!
Lessons: like the retaining clips, early VF attempts were depressingly ugly and I couldn't see a path to success. But again, with a proper post-mortem to catalog the problems, I realized there was a fix for each. Indeed, v2 fixed v1's problems so completely that it jumped straight to production quality.
Step 12: Lazarus (Laser Registration)
The case exposes a wood perimeter, but top and bottom are each covered with a sandwich of acrylic and photo paper cut on Techshop's Universal laser cutter. Since acrylic is the laser's forte, I had no problems cutting the clear windows and their white surrounds.
Cutting the photo paper was trickier. Cutting within and around an existing pattern requires exact X/Y registration between the laser head and image. This was complicated by a DPI variation between the printer and laser cutter which slightly queered scaling on the longer Y axis. The best correction I could make (short of a tricky upscale of the source print), was to use a test pattern that also spanned the axis. This averaged the error down to "acceptable".
Lessons: for precise work, don't trust the laser's sighting dot--assume an error of 0.1" (no problem for blank material). Rather, I started with grids of small squares (1 cm) in the upper and lower spoilage areas. These were successively cut in upper/lower pairs, showing calibration errors as the offset between a square's edge and its cutout. Tweak, rinse and repeat.
Also, to prevent scorch marks on the back side of the paper (from heating of the support honeycomb) I made a lasering jig that floated Bowie in mid-air during the cut.
Step 13: Loving the Alien (The Final "Oh No")
The penultimate step was rounding the outermost edges on the table router. To avoid scratching the plastics, I left the outermost layers of adhesive paper intact. This avoided scratches, but also meant I couldn't see the total case until I the paper came off.
Nervously, I peeled a corner to reveal ... wtf?! The white plastic wasn't opaque! Aaarrrgh! Quite the opposite: plenty of light was leaking through to outline the wooden frame inside. What an idiot--I'd assumed without question that the whites were opaque! Their translucency was a total surprise ... but I soon grew to like it. In hindsight, this things exists becauseof technology so it feels right to glimpse the ghost in the machine.
Lessons:love whatever you make. Yeah it's imperfect but it taught you something and you're better for it.