Sculptural plywood vases

For a second rotary endeavor, I made vases more challenging than carving a heightfield at Plywood noise vases but less involved than a side or coffee table at Sculptural plywood coffee table

Like before, Rhino was used to create the 3D models. Grasshopper and Rhino were used to chop up the models into 3D slices and 2D guidelines. Donuts were cut out for the top and bottom portions of the vases on the normal XYZ Avid table.

Blocks of plywood were glued to make leg stock, but each leg stock had two pieces (one on each end) which had asymmetrical drilled dowel holes. This caused additional attention (and errors) to make sure the dowel alignment was correctly placed within the rotary alignment. UN-intuitively, VCarve places rotary jobs’ A 0 axis as the bottom of the stock when mounted in the machine, yet the 3d model faces upwards. So I had to learn (after destroying two stocks) to flip A 0 to 180 after homing the rotary axis and leveling the stock to a particular face.

The Tower vase’s legs were able to be cut using straight rotary jobs (Y mapped to A). Essentially only one issue emerged, which was when the leg geometry bent up near parallel to the endmill, which caused splintering of the plywood veneer due to the direction of the finish toolpath over the veneer direction. This was fixable with sanding and wood filler.

However, the Tulip vase’s legs did not cut properly using a straight rotary job. These legs do not have material through the invisible A-axis, yet the VCarve rotary job left a default structural spine in place. Additionally, there were many leg surfaces that were too radially tangential to the stock center. This caused the roughing endmill to plunge wildly and caused significant grooving.

As such, I dropped rotary toolpaths and tried regular XYZ toolpaths on the tulip legs, rotating four times around the leg. The first issue was that the roughing pass was too powerful when traveling perpendicular to the veneers and splintered off a chunk of the leg stock near the live end. This effectively separated the main stock from the part held by the live center, meaning that the main stock lost its supporting end and trashed the part.

Solving that required drastically slowing down the feedrate and pass depth plus changing the direction of the toolpath to run along the veneers. Generally, running toolpaths parallel to the veneers causes splintering and blows out the plywood, but in this case it was a better solution than cutting perpendicularly. The roughing pass was further split up so that the tail end was roughed first, slowly, at 0 and 180 degree A-axis rotation. Once the tail stock was roughed yet still connected, the roughing feedrate could be sped up for the 75% of the stock closer to the chuck.
Another issue was bounce. As the tail end of the stock is only held up by a single point, cutting XYZ at a corner near that tail end means that there is no support underneath the material. The endmill effectively puts moment force onto the part during plunge moves, causing bounce. Reducing the plunge rate was necessary again.

On top of that, in order not to destroy the fine border edges during an XY cut by having the endmill drop in Z too far down (such as to the bottom of the stock envelope) and plunging the spindle into the material, I had to make custom roughing 3D models for each leg with thin fins around the perimeter. This prevented the software from plunging down to Z 0 (the bottom of the stock) and instead cut a two-sided, custom-bottom-height roughing shape for each leg.

The finish passes were not cut on normal 0, 90, 180, 270 A-axis angles. Draft angle analysis was done within Rhino to find the best angles for each face of the leg geometry. One leg had its best angles as 5, 65, 185, 305. Yet this required four to five separate files for each leg using VCarve, not Fusion.

Finish passes were also run in XYZ but along a 45-degree angle. It also became necessary to break up the finish passes into two parts because of the geometry. Sharp delicate corners were cut first, when most of the rough stock was still in place. Then the finish toolpath could take a different angle or relocate to a new location to avoid too much blowout of the remaining rough stock.

Once the first leg was cut successfully with separate toolpaths (manually rotating the A-axis within Mach 4 before each subsequent toolpath), I combined the G-code rotations all into one file. The first crash after this came when I did not have the spindle retract fully above a rotating block of stock. The issue was that when I ran a ordinary rotary job, Z 0 was set to the center of the rotary axis. When I ran a XY job mounted to the rotary, my Z 0 within the VCarve file was set to the bottom of the stock - not the center of the rotary axis. So when the XY job sent the spindle to its safe height (say, 1/2" above the top of the XY stock), that was not high enough when the stock is wider than it is tall. The wide stock, rotating, crashed into the stored endmill because the stock rotates higher than the safe Z height. So I had to calculate the radial dimension of the XY block of material, add half the Z block height, then set the safe Z height 1/2" above that number. I also had to manually lower Z 0 within Mach 4 because the touch plate will set Z 0 to the rotary axis for every tool change between roughing and finishing. Yet another learning experience.

After all that came a lot of sanding.

For more finished pictures, see and



Excellent- and much appreciated -explainer. I wonder: you used Fusion for draft angle analysis? Why not for CAM?

1 Like

I used Rhino for the draft angle analysis, not Fusion.

Beautiful work! Your tables on your website are incredible as well, I would love to see more of your process on those.

1 Like

This is absolutely incredible work!! I love the laminate plywood look.

Curious: Have you ever tried Fusion for this type of work? I’ve been messing around with doing full 4 axis work with it a while back:

While my demo was a little silly, it proved out that I could do a toolpath to keep a ballnose “normal” to the surface (And limit the max tilt) pretty easily. This is something you can’t do in Aspire because of the “turkey on a spit” way they do it there, Fusion really let’s you use XYZ and A at the same time.

I was able to use some other toolpaths to do round corners and do general indexing.

I honestly haven’t tried it on something like your design, I wonder if you have?

Eric, are you using the basic Fusion CAM or the advanced features bundled with the Machining Extension to achieve that toolpath?

I am using the machining extension which unlocks a lot of this stuff.


No, I haven’t tried Fusion yet. 3+1 machining took care of this project; 4-axis wasn’t necessary. Since I own a license of VCarve, I am not inclined to pay to rent Fusion unless it’s necessary. At that point, I might look into RhinoCAM so I can keep all my work in one primary software.

I have seen people do some incredible rotary work with RhinoCAM, I’ve never tried it myself. I suspect that it might be good for exactly what you are doing.

Awesome job, Matt! Thanks for the write up and photos.