Nitinol Animatronic Eye

This Animatronic Eye Mechanism uses nitinol memory wires for simple and silent actuation. Custom PCBs enable precise control and feedback for each wire’s position. 3D-printing the mechanism's components has allowed for quick iterative prototyping.

Quick project summary:

Step 1: Defining the project scope and goals

Step 2: Background research and ideation

Step 3: Developing the nitinol control PCB

Step 4: Iteratively prototyping the eyeball mechanism

Technical skills employed:

Hardware: Flexinol SMA wire, custom-designed nitinol control PCB, Adafruit PCA9685 PWM breakout, Adafruit ADS1015 ADC breakout, Adafruit Metro Mini

Prototyping: SMT reflow soldering, electronics component selection, PCB design and procurement, Design for 3D printing & assembly

CAD: AutoDesk Fusion 360 - parametric modeling, assembly modeling, circuit diagrams, PCB modeling

Fig 1. The cheap servos used in my previous animatronic mask were much too noisy and undermined the overall illusion

Defining project scope and goals:

I had a ton of fun creating my first animatronic mask a few years ago and have wanted to build a more advanced version. The biggest challenge has been finding affordable, silent servo motors, as the mechanical whir of even the most high-end servos undermines the illusion of animatronics when up close.

I aim to create truly silent animatronics that don’t require expensive components. I’ve decided to start with an eyeball since it’s a natural addition to my mask project and requires 3-dimensional rotational motion, making it an interesting challenge.

I have defined the scope of this design as follows:

  • Mechanical actuation is inaudible at close proximity (~1 ft)

  • Costs less than $25 USD for actuator and mechanism components (estimated based on portions of materials used)

  • Minimum +/- 35 degrees of motion up/down and left/right

  • Can be powered using commonly available batteries or power supplies

  • Designed with ease of assembly and maintenance in mind

Fig 2. This snapshot shows an early iteration of my test setup. I’m using v1.0 of the nitinol control PCB with two opposing wires.

Background research and ideation:

To begin the process of designing the system for controlling my nitinol actuators, I scoured the internet for information on nitinol’s capabilities and applications, including existing control circuits. I gathered the following information:

  • Nitinol is a 50% nickel, 50% titanium shape memory alloy. It is flexible at low temperatures when in its martensitic phase (characterized by a more easily deformed crystalline structure). At higher temperatures, the alloy converts to its austenitic phase (rigid, compact crystalline structure), causing it to “remember” and return to a predefined shape.

  • Commonly used as an actuator in wire form, with 3%-5% axial deflection.

  • Control circuits heat nitinol wire by passing a current through it, requiring a reliable, constant current source to ensure the wire is not damaged.

  • PWM signals control wire length proportional to duty cycle and facilitate quicker, consistent heating.

  • Actuation time depends on heating/cooling speed, sometimes resulting in considerable hysteresis. Using opposing wires reduces hysteresis for finer control.

  • Nitinol wire resistance is proportional to length. If measured accurately, it can be used for positional feedback.

Using this information, I developed a control circuit as shown below:

  • Breadboard Mockup

    Breadboard Mockup

    I first constructed this rough mockup of the nitinol control circuit using electronics components that I had on hand. This was mostly a proof of concept, and was definitely not optimized.

  • Perfboard Prototype

    Perfboard Prototype

    Once I felt confident that the circuit could work, I started locking in design variables, such as actuation voltage and nitinol wire length. I then sourced the necessary components for a more permanent prototype online, and soldered it together on perfboard. After testing and validating this more robust version, I moved on to modeling the circuit using Fusion 360.

  • PCB v1.0 Schematic

    PCB v1.0 Schematic

    Using Fusion 360's ECAD features, I was able to develop a PCB of my custom circuit. I began by selecting my components and defining the electrical schematic.

  • PCB v1.0 Layout

    PCB v1.0 Layout

    After defining the schematic, I moved on to the PCB layout, including component placement, defining traces and ground fills, and silkscreen text.

  • PCB v1.0 Model

    PCB v1.0 Model

    Although not strictly necessary, I also created a CAD model of my PCB in case I ever want to integrate it into a larger assembly project.

  • PCB v1.0 Prototype

    PCB v1.0 Prototype

    Once I confirmed that my PCB design conformed to the design rule checks (DRC) of my chosen fabrication service, I sent off my files. A few weeks later, I received a set of manufactured PCBs. After integrating and testing the nitinol actuator, I concluded that the position feedback signal needed significant refinement to be useful for closed-loop control.

  • PCB v2.0 Layout

    PCB v2.0 Layout

    I updated my circuit to include additional components to refine the position feedback signal. While I was at it, I also converted the existing components to SMT (surface mount) versions to increase the compactness of the design - except the resistor R1, which I want to be able to easily desolder and swap for different values to accommodate different lengths of nitinol wire.

  • PCB v2.0 Prototype

    PCB v2.0 Prototype

    After testing my updated PCB prototype, I've seen remarkable improvements in the position feedback signal. I also discovered some areas that needed more fine-tuning, as shown in my v2.2 PCB layout.

  • PCB v2.2 Layout

    PCB v2.2 Layout

    In the most recent iteration of my design, I added a new input pin for an external reference voltage and some more components to smooth out the linear voltage regulator and feedback signal.

  • PCB v2.2 Test

    PCB v2.2 Test

    Although there is still room to fine-tune some of the passive component values for an even smoother feedback signal, the current PCB v2.2 is a fully functional prototype. On top of the PWM input and position feedback output, it also features a small potentiometer which can be used to adjust the “zero” value of the position feedback signal.

Here's an example of the nitinol positioning without improved feedback for closed-loop control. The resolution of the position measurements was too low to be useful for fine controls, but the updated circuit should fix this issue!

Fig 3. An early concept involved an outer socket holding the eyeball, with wires pulling from inside. There was too much friction for this to be practical.

Iteratively prototyping the eyeball mechanism

I’ve been developing the mechanism for my animatronic eyeball synchronously with the nitinol control PCB so that updates and improvements to either component inform design choices for the other.

Many existing designs for animatronic eyeballs are available online, but all of them use servo motors for actuation. I want to see if I can come up with a mechanism specially suited to nitinol actuation, which means there are several interesting design constraints.

  1. The small range of motion of nitinol wire requires mechanical amplification through levers or something similar.

  2. Nitinol wire can contract itself under power but must be stretched back out with some external force. It will not stay contracted under tension once power is shut off. To save energy, the eyeball mechanism must be able to hold its position without active control by the nitinol.

I’ve used 3D printing to rapidly prototype my design ideas. This has been an invaluable tool, but it has also presented some design constraints based on the material properties of 3D-printed parts. Primarily, the roughness of 3D-printed surfaces has made friction a large factor in my designs.

You can see the progression of my mechanism design below:

  • Initial Design Concept

    This design concept was heavily inspired by existing mechanisms available on the internet, such as Will Cogley's 3D-printed animatronic eye assembly. I quickly realized that this mechanism is better suited to rigid linkages and servo motors than to flexible nitinol actuators. Reflecting this, I determined to come up with a concept tailored to nitinol's advantages and disadvantages - strong and flexible, but only able to supply tension and with a small range of actuation.

  • Second Design Concept

    My second design concept for the eyeball mechanism consisted of a hollow eyeball encased in a socket. The nitinol wires attached to a smaller sphere within the eyeball, magnifying the wires’ relatively small range of motion. The idea was to exploit the simplicity of nitinol as an actuator by minimizing linkages or other moving parts.

  • Iterating through Potential Solutions

    Upon testing my initial design solution, it became clear that the biggest drawback was friction between the 3D-printed parts. After iterating through a dozen or so prototypes - varying tolerances, surface finishes, and lubrication methods - with little success, I decided to change tack to a new design solution.

  • Second Round Ideation (1/4)

    After returning to the drawing board, I brainstormed as many novel joint concepts as I could, even if I wasn't sure they would work. I made rough initial prototypes for a few, including this one that uses four cords in tension.

  • Second Round Ideation (2/4)

    This concept uses a much smaller internal ball joint - the idea was that the much smaller surface area would reduce friction considerably, but the design was also (predictably) much too weak and easily broken.

  • Second Round Ideation (3/4)

    Second Round Ideation (3/4)

    This concept uses a cord in tension, running axially through the cross-shaped rocker that holds the eyeball. I was able to get this one working, but it wasn't easy to assemble.

  • Second Round Ideation (4/4)

    Second Round Ideation (4/4)

    This version uses a pseudo-ball joint inspired by the mechanism inside a joystick. I was struck by how similar the motion of a joystick is to the desired motion of an animatronic eye. This solution is reliable and easy to assemble, making it the top contender. I'm still tinkering with similar solutions to see whether I can improve it.

Next

Zephyr Kite Steering