Fabricating Flat-pack Furniture - First Foray

This document follows the process I used to create a 3 legged stool. The project needed to fulfill several basic requirements:

1. I bought a new CNC machine and I needed to make a large cut to prove it functional.
2. The CNC machine is difficult to mount on high anchor points. I needed a taller platform to stage the machine for calibrations and mounting and dismounting.
3. I needed to develop a reliable procedure for modeling and fabricating flat-pack furniture. This first project would provide insight regarding CNC-specific issues like:
4. What tolerance do I need to add to slots for them to be tight enough to add structure but loose enough to assemble easily?
5. What software is most capable, flexible, and free? (And allows you to be both precise and creative?)
6. What file formats does the machine need? (And what software can convert a novel design into that format?)
7. What settings in a cutting job result in the best parts for this machine? (What feed rate works best for the medium? What tab sizes sufficiently hold the parts but are removed easily afterwards? What is the maximum cutting depth the machine can make? How well can I set the machine's starting point? How much distance do I need to put between my cut parts?)

For these reasons, I chose to make a 3 legged stool. I thought it might be simple enough to make with minimal pieces and it could be made to dimensions that would actually be useful later.

Conception

1. Computer (first for designing and then for CNC control)
2. OSB/Plywood Sheet (I used 4'x8'x1/2" OSB)
3. CNC Machine (& router & bit & dust collection)

Assembly

1. Jigsaw or Pull saw (for part cleanup)
2. Oscillating Blade (for cutting tabs and removing parts)
3. Soft Mallet (for persuading tightly fitting parts together)
4. Wood glue (to strengthen assembly - optional)
5. Paint (to protect result - optional)

I took a tape measure to the CNC router's frame and to the machine's sled to help decide upon broad measurements.

1. The sled of my CNC router has a 16" diameter.
2. The OSB sheet I'll be cutting has a nominal thickness of 1/2".
3. The lowest the stool could be and still be helpful during the mounting process is 3' tall.

I also decided that the strength of the stool would be improved by restricting the legs movement with a ring. I thought the same could be done on the inside (and this could even pull double-duty as a storage area.) The outer ring could pull double-duty as a footrest. The inner disc could be designed to sit a little lower and the shelf would even have a built-in "lip."

1. I decided upon a footrest height of 1'.

Commercial CNC machines are expensive! Especially large-scale machines. ("Should I buy a CNC machine or a new car?")

But if you want a CNC and can't afford a new one, don't want to buy a broken one on Craigslist for nearly the same price, or manufacture a machine from scratch yourself, then I think there are 2 main options: a Maslow CNC or a LowRider CNC. Either one can probably be purchased and constructed for less than $800 (depending on frame/table, router, and bit choices.)

I bought a Maslow CNC 4 Kit and then a 4.1 upgrade kit. And then I never found the time to put it together for a year... (In my opinion, the tradeoff for these kits is that although they're much cheaper, there's more you have to do yourself to get them up and running and there's more you have to do to maintain and repair them.)

When I finally had time over holiday break to put the CNC machine together, I found that one of the pieces was incompatible with the upgrade kit. After an excellent customer service experience and waiting a few days for shipping, I had all the necessary parts and I couldn't put it off any longer. So I assembled the machine:

This CNC uses 4 anchor points and belt motors around a central router spindle instead of mounting the router spindle to a table & rails & a gantry like most CNC machines. I like to compare it to the NFL Skycam. (Except that the Skycam places the motors and reels of cable at the anchor points instead of on the machine's sled.) But due to this movement method, there's more flexibility in what the machine can cut. (You can cut slightly warped sheets for example because the sled will slide over the top surface it's cutting.) But it also means that we need to create rigid anchor points so that it can compute its location and movement accurately.

The Skycam layout also means that the Maslow can work in many different orientations (so long as when it moves, it's held against the workpiece.) One of the popular choices is to use the machine in a horizontal orientation and drill holes for anchor points in the concrete floor of your shop. I wanted to be able to cut full sized sheets in my garage (because I didn't want the noise and dust in my house) but my garage is single-car narrow. The geometry of this machine's technique means that the anchor points need to be outside the dimensions of the cutting piece. So I opted to create a "default" frame in "vertical" orientation.

I added additional corner bracing, two homemade gate kits, and threaded rods for anchor bolts to my frame. (With threaded rods, I hoped to raise each anchor point to match the z-axis "height" of each motor arm - they're stacked on top of each other but that means they're pulling at different angles.) I hope to later add hinges to the top and bracing at the bottom so that it can easily "fold" into the wall. But this frame is really rigid and really square already so I'm happy to start using the frame before making those improvements.

I also purchased a power strip and an emergency stop button so that when something goes wrong I can interrupt the CNC machine, the router, and the dust collection with a single slap.

I also found my wet/dry vac and printed a dust shroud and adapter to fit my vacuum hose.

I love Blender and will try BlenderCAD or Blender CAD Sketcher eventually but I thought for my first project at least, I’d use FreeCAD or Fusion 360 personal so I wasn’t immediately deviating from an accepted workflow.

The good news is, I didn’t have too much trouble designing my stool in either FreeCAD or Fusion. They both use constraints in relatively intuitive ways and allow you to export DXF files from sketches. FreeCAD is free and Fusion 360 personal is free for non-commercial projects.

Constraints are geometric rules or restrictions on shapes in the sketch. The curves on the inside of the legs are made with control points dragged by mouse until aesthetically pleasing but every other line in the sketch is defined with at least one constraint. (Most of the lines have 2 constraints: a distance constraint that constrains the length of the line, and another that restricts its orientation to either a horizontal or vertical direction.)

An extra note: I measured the thickness of my OSB sheet and although it was far thinner than a 5/8" sheet would be, it was also far thicker than 1/2" (and especially thicker than the .483" actual thickness it claimed to be.) I made a huge mistake and drew the slots and tabs as if what was going to slot into them was going to be 12.7mm thick. When I do my next cut, I'm going to trust my precutting measurements for the slot widths and I intend to add .3mm or .4mm for additional tolerance.

Exporting from sketches in both FreeCAD and Fusion 360 to the DXF file format is easy. I didn't find any immediate issues with the default settings in both software. Planning the job requires an intermediate (online) tool step. But the first step is to export the DXF from the sketch.

Once each part has its own DXF file, I used KrabzCAM to arrange them into a cutting plan. At first, I had them arranged in a compact enough plan that they'd only use a little more than a half-sized sheet. But I didn't feel confident about how much material I needed to leave outside/between each part or how well I'd be able to position the CNC machine so it didn't cut off the edge.

Because I was using a 1/4" single flute bit in OSB, I chose the fastest feed rate I was comfortable with and the slowest spindle speed. I also added tabs and dog bones. (Tabs are small sections of material left uncut that prevent cut or nearly cut parts from falling or moving before the job is finished. Dog bones are additional little "bump ins" on paths that improve inner corners.)

I selected each shape that I wanted to cut at a time and created toolpaths for each. (In media this thick, you have to cut shapes in several shallower passes. You can either cut all the way through on the first leg and then continue onto the other legs, or you could cut all the first layers for all three legs and then cut the second layer for all three legs and so on.)

Then I simply saved the G-Code. This will prompt your browser to download a .NC file. This is the file you'll need to get onto your CNC controlling computer.

But before you run your cutting job, you may want to sanity check your work.

To estimate how long the job would take, I opened the NC file and copied the GCode instructions into Nicolas Raynoud’s Q’n’dirty Toolpath Simulator.

To see whether the cuts and tabs looked appropriate "to my eye," I opened the NC file in CAMotics, created a 1/4" tool, and simulated the cut.

Both of these tools also gave me an estimate on the bounds of the job. (Because I'd be cutting parts with a bounding size far below the size of my starting sheet size, I knew that I could be off by several inches in setting my home position. This gave me more confidence in initiating my first cut with my new machine.)

Finally, I couldn't delay the inevitable any longer so I started my first cut!

I mounted the machine, had to recalibrate it a few times, uploaded the NC file, set my home position and Z-axis home, and then hit "Play."

It worked!!!

... for a while. Then it halted partway through cutting the third leg. I hit the emergency stop, knowing that I'd have to finish cutting that part out by hand.

I quickly created a new job in KrabzCAM (forgetting to export tabs and dog bones this time), reset the home position for the unspoiled right side of the sheet, and hit play again.

Because I forgot to add tabs, when it finished cutting the inner disc, that part fell to the bottom of its own cutting path, and then started spinning on the still running bit. I hit the emergency button again, knowing I'd have to finish the third leg and the outer ring by hand.

Removing the parts from the sheets was pretty annoying. This is a step I expect to improve greatly with more experience. I’ll have better material next time. I’ll carefully measure the actual thickness next time. I’ll babysit my laptop and the machine better next time (and probably tell my laptop to forget all its other wifi networks and never sleep.) And I’ll have sharp jigsaw blades, an oscillating saw, and a pull saw ready for this step.

This time though I only had mismatched saw horses, a dull jigsaw blade, a sawzall, and an oscillating saw. It was cold enough my fingers hurt. And I had several parts that weren’t completely carved out by the CNC.

I brought the extracted parts to the warmth of the basement so that the wood glue I used would set and cure. (When I get better at creating tight but tolerant fitment, I wonder whether I'll still want to glue flat-pack parts together?)

I also used a vice and files to “smoosh” and sand the tabs closer to the nominal thickness so they'd fit in the narrow slots.

I spread wood glue by finger on all the tabs and the parts that would meet with the flat of the discs or ring.

Then I used a soft mallet and eventually a sledge hammer to persuade the parts together. (The hand finished third leg was especially difficult.)

Except for the job not completing (and needing to be recalibrated even though it had already been calibrated on this frame) this machine is amazing! The difficulty in putting my stool together was due to trusting supposed material thickness over my measurements, and finishing by hand a part that was only partially cut by the Maslow.

And my design has a flaw. When you step on the ring to sit on the stool, too much of your weight is beyond the supported area of the stool. (It tips.) I could have redesigned the ring (maybe made it triangular?) or given the stool 4 legs so less of the ring would extend beyond its footprint. 4 legs wouldn’t have even had to wobble...

I may still add 3 additional legs at the halfway points between each current foot on the ring. I think a six legged stool sounds dumb but I’d like this first project to still be… sit-able.

In the meantime, I think it’ll work great for a Maslow pre-mounting stand!

My Summary:

After spending less than $1,000 on a new CNC machine, router, bit, some additional electronics, and 3d printing some adapters I can cut computer-aided designs with about .5mm accuracy from a full-sized sheet of lumber up to 50mm thick! I count the entire project a success. (Compared to most of my other projects this was remarkably minimally buggy.)