Perpetua Stool

The Perpetua Stool combines traditional woodworking tools with 3D modeling software and 5-axis CNC machining to create forms through a digital workflow that are as repeatable as they are sensual. This project matches the sensitivity to detail hand-carving techniques with the efficiency and geometric accuracy of digital fabrication. It's part of a personal quest to bridge the distance between old and new.

The stool is a collaboration between Matthew Sellens - a Portland, OR furniture maker and owner of Perpetua Wood Floors - and Autodesk Pier 9 artist and CNC researcher Xander Bremer (that's me). The design was developed in Portland where it moved through numerous forms as it passed from sketch to prototypes. With his decades of experience in traditional woodworking, Matthew brought an efficiency and intuition to the manual design and iteration process, and I was the one providing feedback and doing the digital aspects of the design, and getting the project streamlined for CNC fabrication.

The design leverages every aspect of hand work - sensual curves and subtle details that can arise only from touching the wood with one's hands. It is well-proportioned and ergonomic, which are hard to achieve when working on sculpted geometry in a purely in the digital realm.

At the same time, it fully leverages the range of DMS router and a suite of Autodesk 3D modeling software. It has the fully sculpted contours that digital fabrication is so excellent at producing, as well as using the 5-axis capabilities to accommodate high-precision, compound angle joinery all incredibly tightly referenced to each other. Both of these make details possible that would be incredibly painstaking to produce purely using traditional tools.

The rough steps of the process are shown here. Hope you enjoy the peek into this process.

More of my work is on my website as well.

The design of the stool took about a year of development as a side project to Matthew's wood floor business.

The process started with sketches and hand-drafted layout of the stool, which became the basis for templates for roughing out the main profiles. Mattie allowed the design to wiggle through different iterations as it addressed challenges around stability as he sought a clean profile that expressed an intelligent, organic minimum of material to create a strong stool.

In the photos you can see the laborious process Mattie undertook, with each part being sculpted by hand, before arriving at the final design. I had to do the hard work of sitting on my butt and responding to photos that came my way, providing a little feedback and thoughts on CNC constraints, while he worked on his blisters.

Matthew and his partner in business and crime (aka Gloria - hi mom!) used Autodesk's 123D Catch to create a 3D file of the the seat and legs of the stool that they sent to me here in San Francisco.

I used a process I had developed in earlier projects at Pier 9 to take mesh files from 123D Catch through Meshmixer, Netfabb, and Mesh Enabler to get solid parts in Inventor to use for the basis for CNC toolpathing in HSMWorks. I wrote an Instructable last fall about that workflow.

With the toolpathing done and prototype joinery and surfacing strategies complete, the team assembled in San Francisco so Glo and Mattie could watch me execute the first CNC produced part. Mattie stared transfixed as the DMS made quick work out of what would have otherwise been days of careful sculpting. We were able to discuss right there at the machine as we saw things that worked well, and others that would need more work in every phase - the model, the capturing, the CAD and the toolpathing before we were ready for the first full version.

Back in Portland, Matthew and Gloria had the stool 3D captured by a pro (Danny, at Global Inspection Solutions) so that the mesh was in no way a limiting factor keeping us from getting the perfect shapes. The resolution on the scans was so high he had to send lower resolution files, which I even then had to dumb down just a little more to keep the triangle count low enough for Inventor. But the accuracy of this digital geometry was incredibly good.

I went through the same workflow described in the Instructable linked earlier, and all parts were checked for scale to make sure they matched the first cut prototype, which matched the original.

Once in CAD, I designed all the joinery, toolpathing, stock material, and workholding that would make it possible to extract the final piece from the digital world and make it physical. Some of the details were incredibly intricate, to reference the different parts to fit together properly. Without the software and machines to allow these precise compound angles, it would have been quite a tricky task.

The major trick for me was to make sure that I cut the models in the appropriate places so that when everything went back together, it would be perfectly flush. It was crucial to define hard references on each mesh to establish a strict working surface to then constrain the different parts to each other. I tried to show some of these key cuts and references in the pictures for this step, which also show some of the supporting sketches, planes, and mirrors that made it all happen.

The cross bar was the last part modelled, and the only part of the design that was not 3D captured. I made a simple model that left a little extra material to ensure that I could remove the last little bits to be smoothed by hand to match the rest of the contours.

For the seat, it required a part flip with roughing, finishing, and joinery on the bottom, and then roughing and finishing passes on the top. I used a web around the edge of the flat edges of the model to hold onto the part after the flip, and a slot in the bed of the machine to orient the part on the flip.

A small stepover with a 1" ball end mill left me a very smooth surface ready for hand-finishing. I left the part in the web so I could easily clamp it to the table for sanding and assembly.

The legs were a complicated beast.

I needed to do some complex 3+2 machining at some weird compound angles, which meant the head of the machine would be put well below the part. In order to provide clearance for these operations, I had to make risers to hold my work high in the air, which then also had to be made square for part alignment, include vise bolts, slots for vise dogs, optional clamping bolts, and flanges to clamp the work to the table of the machine.

Hours of work just went into the workholding on this part, and so it was a relief when it came time to "just" make the part.

I started with the rear legs, which only had critical dimensions on one end. I cut the rough surface of the leg with a large flat end mill, which then pivoted to cut the joinery at complex compound angles (you can see B and C in the machine controller photos, which are rotations about the X and Z axes of the machine). Then I smoothed off the contours with a ball end mill, and added joinery details with a small flat end mill. This way, after flipping the part, all I had to worry about was creating the other contour without machining through a .15" flange that went all the way around the part to hold it stable in the machine. There was no risk of an error in fixturing affecting the precise joinery - a surface with a slight defect is easily sanded flush - a joint, less so!

The front legs followed the same process, except that there was an additional compound joinery detail on the inside face of the legs for the footrest.

The crossbar was the last piece made, and was a little tricky because of its small size and highly contoured profile. But I used small flat flanges to hold it, which I could easily remove later.

With all the hard work done, all that was left was assembly, trimming away the workholding, and sanding everything to be flush and smooth to the touch.

Thanks to the precision tools, there was not too much hard work left at this point in the process. The workholding flanges came off quickly with a carefully adjusted spokeshave, and the joinery all fit impeccably, since I tested the cutting radius of the end mills I was using during another project I was working on at the same time.

There were a couple small adjustments to contours where they didn't match up perfectly. This was likely due to a combination of lowering the mesh on the 3D capture, cutting joinery, and then just normal slight deviations from machining and the wood moving a little bit (as wood does). These details will be changed before the next part is machined, but it was still easy enough to correct by tracing lines to hit with curved gouges and rasps.

The clamp up required a little brain power, but yielded excellent results and a fun problem solving opportunity (look closely!).

From there, it was just a few hours of scraping and sanding up to 400 grit before applying a couple coats of linseed oil.

The Perpetua Stool was an amazing opportunity to combine concepts of old and new - organic contours and precision geometry; Matthew's decades of woodworking experience with my expertise in digital fabrication; handmade and machine made; and collaboration that is both intimate and remote.

It was such a privilege to share this experience with family. I really have to give a shoutout here to Autodesk/Instructables/Pier 9 for making this project a possibility.

It's such an exciting thing to be a part of a community creating new ways of collaborating, and thinking creatively about new opportunities that opens to combine different skillsets. To have Matthew's sensitivity to wood design so deeply embedded in the project was such an inspiring starting point as a collaborator to add additional value to the project.

I hope it's a sign of things to come - not just for our family. But for how these tools, as they become more accessible, may change the way we are able to work: intimately and remotely, alone yet together, and differently, all on one thing.

Thanks for taking a look.