Build Your Own Drone

I chose a multi-rotor aircraft over a fixed-wing or helicopter because I wanted to be able to get a static view. Multi-rotor aircraft are also fast, very stable, and able to launch vertically in the tightest of confines—even inside a room and out a window.

Multi-rotors typically use anything from three motor-propeller assemblies to eight, mounted at the end of arms that are centrally interconnected. I chose the “Y6” configuration, composed of three motor-mount arms with two co-axially mounted motor-propeller assemblies at the end of each arm. With two motors per thrust point (one facing up as a “tractor,” and one down as a “pusher”), the Y6 has redundancy. Because it has only three arms, it gives a mounted camera a wide field of view. I needed the UAV to be as small and stable as possible. Smaller fixed-wing UAVs often fly “squirrelly,” but that’s not true of a well-designed Y6, with its multiple points of thrust stabilized by a high-performance flight control computer.

 

Multi-rotors fly with remarkable speed and stability because they have a number of points of thrust, not just one. Each works against and with other thrust points—and with and against gravity—to move the craft along three axes, and, if needed, hold it steady in one position.

A multi-rotor pitches, rolls, yaws, and hovers by varying the speed of its motors (each connected to a fixed-pitch propeller) individually, which varies thrust (for pitching and rolling) and torque (for yawing). This type of aircraft, however, is inherently unstable unless “balanced” by a very powerful flight control computer, one that can analyze aircraft attitude and position, then provide control inputs (as motor speed rate changes) orders of magnitude faster than a human’s ability. Think of trying to balance a baseball atop the tip of a pencil: You’re not really “balancing” it, but constantly moving the pencil under the baseball in a dance with gravity to get a few brief moments of relative stability. But most people don’t have the eye-hand coordination for such a feat. Similarly, until recently, sensors and computers simply couldn’t work fast enough to use multiple thrust points to control a small aerial vehicle.

Over the past few years, the electronics industry has made great strides in the development of micro-electromechanical systems and inertial measurement units. These include tiny, solid-state, multi-axis gyroscopes for spatial orientation and accelerometers to measure change in velocity to guide multi-rotor and other types of aircraft. Manufacturers also produce micro-electromechanical magnetometers for navigation, and pressure sensors (barometers) for altitude determination.

 

I settled on a company at the forefront of the technology, Hoverfly Technologies. Their HoverflyPRO control module uses 16 parallel processors in its flight control computer to analyze thousands of inputs per second from the onboard three-axis gyroscope, three-axis accelerometer, and digital pressure sensor.

The controller, a printed circuit board that measures just 2.75 by 2.75 inches by 0.5 inch high, takes flight control inputs from a digital receiver (taking commands from a user-controlled transmitter on the ground) and tells the Kestrel to go, stop, and hover.

The board commands the camera to pivot up and down, and side to side, has an altitude-hold function, and overlays vital flight data on live video fed to a ground station—if a video transmission system is mounted to the craft.

I also bought the HoverflyGPS control unit, which, when mated to the PRO board, adds three-dimensional position hold, automatic return-to-home, and waypoint navigation.

I could have simply mounted all of the components on a pre-built hobbyist multi-rotor body, but I wanted the final aircraft to be as light and small as possible, very strong, built specifically around my components, and have no parts such as wires or electronic speed controllers (ESCs) dangling in the open.

With the dimensions of all of the components, I used illustration software to sketch out the smallest craft possible.

For construction material, I chose carbon fiber (which I ordered from DragonPlate, in Elbridge, New York). It’s extremely light and strong, although notoriously difficult to cut, and it’s an electrical conductor, so I could not have any exposed wires. Carbon fiber is also excellent at dampening vibrations, important for any flight controller, as vibrations can affect the performance of the accelerometer and gyroscope, not to mention the video. Then came the hard part: I had to engineer each individual piece of the UAV body, but I had no experience in engineering or computer-aided design. Most of the professional CAD programs cost thousands of dollars, far out of my price range. I bought an older version of TurboCAD I found on Amazon for less than $30.

Using exact dimensions of the HoverflyPRO controller, ESCs, motors, and so on, I engineered 29 parts in one month.

A caliper micrometer with one of the extruded carbon fiber parts. (Ed Darack)

Virtually constructing the body of the craft with TurboCAD, I was able to include folding-forward arms for ease of transport, perfectly align all holes for screws (I wanted everything to be connected mechanically, with no glued parts), and gain a sense of how it would balance with all components mounted—vital for multi-rotors.

With the parts engineered, I needed them milled from the pieces of carbon fiber stock. This turned out to be much more difficult than I imagined. Few machine shops work with carbon fiber, as it “eats” cutting bits. I finally found Jason Sauer of Pinnacle Machining in Fort Collins, Colorado, who agreed to help me.

There is no universal file format for CAD; Sauer was quickly able to redraw the TurboCAD files for the simple parts by hand, but for the more complex pieces, I had to figure out how to convert the files.

Stymied, I put an ad on Craigslist, and within 30 minutes heard from Tom Hanson, a machinist-turned-engineer who has his own firm, Hardware Collaborative. He was intrigued by my project (he often donates his time to educational engineering projects), and converted the files in minutes.

With the files completed, Sauer cut all of the parts—to a .0001-inch tolerance—on his three-axis Haas computer numerical-control milling machine.

I then bought an assortment of black anodized hex cap screws from C D Fasteners, and very-hard-to-find black anodized aluminum locking nuts from Fastener Express.

Thanks to Sauer’s skill, the pieces of the craft fit together perfectly. Carefully cutting, soldering, and shrink-tubing (to insulate and protect the soldered wires and connectors), I finished the construction of the Kestrel-6.

Step 6: Moment of truth