Designing a DIY Augmented Reality Headset

Bradley Ramsey
Supplyframe
Published in
5 min readMar 29, 2018

Kelly Peng describes her journey into the world of AR technology and obstacles she faced along the way.

Augmented reality (AR) is an incredible technology that sounds like it was pulled straight from a science fiction novel. At HDDG 28, Kelly Peng spoke on her journey into the realm of AR and the obstacles she faced when building her first prototypes.

Kelly is the founder and electrical/optical engineer of Kura Technology; a company that manufactures augmented reality glass. Her presentation dives deep into this emerging technology, and how you can make your own low-cost AR glass at home.

Kelly’s Next Big Project

Kelly Peng has quite an impressive resume when it comes to her past projects. In the past, she has worked on a deuterium-based nuclear fusion reactor, Raman spectrometer, DIY structured light camera, active EEG sensors, and even like/dislike emotional classifiers.

Somehow, among all of this, she also hosts a hardware hacker and R&D community Inventor Lab which meets for monthly hack sessions. Their current project? A linear particle accelerator that works in Kelly’s garage!

While working on emotional like/dislike classifiers, Kelly realized that there was a missing piece. Kelly realized that people would want to wear it everywhere and receive real-time visual feedback. This train of thought led her to discover the solution: AR glass.

Solving Problems Associated with Optics and Field of View

As Kelly researched the current state of the market, she found that optics were a major obstacle. The other major issue was the field of view (FOV). When she first began conceptualizing her AR solution, Kelly found that most options on the market have a very small FOV between 10–23 degrees.

She looked at what other companies were using for their optics. One option was a diffraction waveguide which could be made thin and small. On the downside, it was hard to manufacture and was limited to an FOV of about 50 degrees.

Another option was a reflective waveguide. This offered less diffraction and ghost imaging, but it was hard to align the layers in mass production. Furthermore, the FOV was once again limited.

She settled on a diffraction wave guide, but she was determined to make it herself to cut costs. Kelly knew that she wanted to create AR glass that had a human-sized FOV. Her goal was set at110 degrees.

To fulfill her goal, Kelly and several of her friends started creating designs using Autodesk Inventor. They wrote their own OpenGL, C++, and Python scripts to generate more detailed parameters and turned the surfaces into .stl files.

Using an open-source rendering software called Blender, Kelly and her team were also able to test their designs without the need to manufacture them.

By using the camera in Blender and a checkerboard texture to simulate the display panel, they could simulate the position of a user’s eye and adjust the shape in Autodesk Inventor accordingly.

The team also wrote an OpenGL C++ script and a rendering engine based on DirectX to accomplish three tasks:

  • Draw the major rays
  • Perform pixel-to-pixel rendering
  • Adjust position for viewpoints

Once they had created their design virtually, the next obstacle was manufacturing a prototype without breaking the bank.

Bringing the Design to Life

The first version of Kelly’s AR headset

As Kelly and her team explored the options for manufacturing, they examined options that included a CNC mold, vacuum forming, and coating with a semi-reflective metallic layer.

Once they had lenses, the next step was to explore options for the electronics. Kelly took LCD and OLED panels from Oculus and Vive headsets. She chose LCD for her panels because of the higher brightness levels over OLED. She sourced her LCD panels from Alibaba and Shenzhen.

During this process, she also made the switch from one panel to two. This would offer a wider FOV if she angled the two panels. The next step was to 3D print the casing for the headset.

The two options she used were Formlab and UPrint, both of which had their pros and cons. Formlab offered a great industrial prototype, but it needed holding screws to handle the weight.

UPrint was fast, but they only offered a white material. The parts also had to be soaked on hydrogen peroxide for two hours to dissolve the supporting material.

The first prototype included these parts inside the headset:

  • A stereo RDGB Camera for SLAM tracking
  • A modified SLAM algorithm
  • Gesture inputs using Leap Motion
  • A Zotac/Qualcomm CPU

Improving The Design For Versions 2 and 3

As the team moved forward, the focus on version 2 was more complexity in the optical design. This began with field sequential LCOS Liquid Crystal on Silicon Microdisplays, combined with a custom designed high current LED driver.

The optical design used in Kelly’s headset design (and many others) is a non-pupil forming design. This offers a larger eye box and a wider range of focus. The electronics were enhanced by using custom boards for driving LOCS and LED.

The new board utilizes a custom ASIC for driving LCOS and PWM for driving the LED light source which are synchronized. All of this is debugged using I2C via Beaglebone Blue.

The goals for version three include shrinking the size of the headset and improving the optical quality. The newest prototype has a resolution of 1080p per eye and runs on a mobile CPU.

The first version of the headset has been turned into an open-source project, while the second version will be leveraged for an optical module product. In the meantime, Kelly and her team will be working to optimize the third version of their remarkable AR headset.

--

--

Bradley Ramsey
Supplyframe

Technical Writer at Supplyframe. Lover of dogs and all things electronic.