Skip to main content

A basic 3DOF acrobatic tail — Build overview

The mechanical build of this tail was designed so that the hand drill I had access to would be the only power tool, along with 3D-printed motor holders from Treatstock. I'm no machinist, whatsoever. If you have the budget, the design is pretty accessible!

The build for this tail prototype could be broken into a few distinct stages:

  1. The first joint
  2. The rest of the skeleton, with the links connected to the second and third joints
  3. The fur and lights
  4. The layout of the base
  5. Integration

It took about 4 months including design time and trial and error. A flawless build of the prototype as it exists today probably would've taken a couple of weeks (so definitely emphasis on the error) and $530 in raw materials according to the BOM. With surplus, iteration, and breakage, I spent $1030 on raw materials.

Download the BOM (.XLSX)

The rest of this post is a description of the way the build components are put together. The details of the trials and errors will be written in the build process post.


This overview of the assembly will be some degrees broader than a step-by-step. I only had so many process photos! Another prototype of this will be much more specific.

Motion hardware

Broadly-speaking, the core moving parts come together likely as you'd expect from a glance.

  1. Joint 0 is a sensored inrunner BLDC joined by a 5mm-1/8" shaft coupler to a spur gear driving a planetary gearbox (Amazon).

    • It is bolted to this gearbox with a 3D-printed PETG mount, which is bolted to the base by two standard NEMA 17 mounts (Amazon).

    • I would suggest (not just to you, to myself too) a replacement of this joint with a dedicated gearbox, especially a right-angle non-worm gearbox aking to these ones on Misumi but with higher gear ratios — non-worm so you can bend this joint by hand; right-angle because this joint is the widest part on the base, since the output shaft needs to be centered widthwise on the base. Further alternatives include motor-gearbox combos for industrial automation, and on the complete opposite end of the spectrum, scrapping a windshield wiper or window driver. This will be updated as I find cheap low-voltage geared brushless actuators.

  2. The 1/4" output of this gearbox drives a flange bolted to the first link. All links are 3/4"x1/16" square hollow aluminum tubes weighing about 3g/linear-cm.

  3. A 180° torsion spring is slotted directly into the first link through a keyhole and secured to the base by being clamped between two aluminum bars with a guide slot cut with a hacksaw. The spring provides 24 kg·cm at full compression (180°), lifting the resting position of the outstretched tail to about 45°.

  4. The first link connects to the second joint — a Hitec D675MW servo — by a 3D-printed PETG mount. The mount braces against the weight of the rest of the tail, which is unidirectional by design of the kinematic chain.

  5. The second joint connects to the second link by a plastic servo horn and two screws: an M2.6 screw through the spline (!) and an M3 through the horn.

  6. The second link connects to the third joint — a 20kg Zoskay servo — through another 3D-printed PETG mount.

  7. The third joint connects to the third link through a metal servo horn with an M3 screw through the spline (!) and M3 screws through the horn.

    • Recommend (myself included) to switch this to a second Hitec D675 (or perhaps slightly weaker): while these cheap servos really do put out some watts, the low-quality gearbox makes this by far the loudest joint when it should be the quietest.
  8. Each link is 19.5cm between each axis of rotation:

    • Link 0 is 18.5cm by shaving a 3cm flange-to-spline distance on the servo, and adding a 2cm flange radius at the joint 0 gearbox output.
    • Link 1 is 24.5cm long, adding 3cm extra for the far-side screws on the joint 2 mount, and 2cm extra for the joint 1 servo horn.
    • Link 2 is 21.5cm long, adding 2cm extra for the joint 2 servo horn.

The joints are arranged in a pitch → yaw → pitch configuration. More details on the kinematics can be found in the control post.


The skin is as it looks other than a hidden inner layer of compartments to keep the stuffing uniform. Broadly:

  1. The fur layer is made up of tapering cylinders of faux fur sewn together — nets of these sections are circle segments. The nets are available for download as 8.5x11 pages (PDF) and as a raw SVG: the right cluster (odd numbers) is gray, the left cluster (even numbers) is brown.

    • The nets were generated by taking the circumferences at each end $c_1, c_2$ and the segment length $l$ and solving the equations:

      $$ \theta r_0 = c_0\\ \theta (r_0 + l) = c_1 $$

      to define a circle segment with inner radius $r_0$, outer radius $r_0 + l$ and angle $\theta$.

    • The gray fur comes from Shannon Fabrics; the brown fur comes from a local textiles shop and equivalents are available on Amazon.

  2. There is an inner tube of fabric from a t-shirt runs the length of the tail and wraps around the skeleton. Velcro strips line this tube in the middle of each link to secure the skeleton to the skin alongside friction.

  3. Between the fur and inner tube is a layer of polyester stuffing. This stuffing is divided lengthwise into three roughly equal compartments by two rings made from t-shirt fabric, to prevent clumping in the tip of the tail. The stuffing is also divided into two compartments radially by fins made from t-shirt fabric, to prevent clumping in the tail underside.

  4. 72 through-hole LEDs are pushed through the fur layer and soldered into a charlieplexed array with 22AWG stranded hookup wire. Schematics are available in the electronics post.

The skin was constructed into its three lengthwise compartments separately, stuffed, then sewn together.

The towel in this photo isn't included in the actual tail skeleton, it was there for protection (both its and mine) when testing the bare skeleton.