When building autonomous airborne vehicles like drones or UAVs, saving a little bit of weight goes a long way, literally. Every gram saved means less energy needed to keep the aircraft aloft and ultimately more time in the air, but unmanned vehicles often need to compromise some on weight in order to carry increased computing abilities. Thankfully this one carries a dizzying quantity of computer power for an absolute minimum of weight, and has some clever design considerations to improve its performance as well.
The advantage of this board compared to other similar offerings is that it is built to host a Raspberry Pi Compute Module 4, while the rest of the flight controllers are separated out onto a single circuit board. This means that the Pi is completely sandboxed from the flight control code, freeing up computing power on the Pi and allowing it to run a UAV-specific OS like OpenHD or RubyFPV. These have a number of valuable tools available for unmanned flight, such as setting up a long range telemetry and camera links. The system itself supports dual HD camera input as well as additional support for other USB devices, and also includes an electronic speed controller mezzanine which has support for quadcopters and fixed wing crafts.
Separating non-critical tasks like cameras and telemetry from the more important flight controls has a number of benefits as well, including improved reliability and simpler software and program design. And with a weight of only 30 grams, it won’t take too much cargo space on most UAVs. While the flight computer is fairly capable of controlling various autonomous aircraft, whether it’s a multi-rotor like a quadcopter or a fixed wing device, you might need a little more computing power if you want to build something more complicated.
If you’ve been following [Joe Barnard]’s rocketry projects for the past few years, you’ll know that one of his primary goals has been to propulsively land a model rocket like SpaceX. Now, 7 years into the rollercoaster journey, he has finally achieved that goal with the latest version of his Scout rocket.
Many things need to come together to launch AND land a rocket on standard hobby-grade solid fuel rocket motors. A core component is stabilization of the rocket during the entire flight, which achieved using a thrust-vectoring control (TVC) mount for the rocket motors and a custom flight computer loaded with carefully tuned guidance software. Until recently, the TVC mounts were 3D printed, but [Joe] upgraded it to machined aluminum to eliminate as much flex and play as possible.
Since solid-fuel rockets can’t technically be throttled, [Joe] originally tried to time the ignition time of the descent motor in such a manner that it would burn out as the rocket touches down. The ignition time and exact thrust numbers simply weren’t repeatable enough, so in his 2020 landing attempts, he achieved some throttling effect by oscillating the TVC side to side, reducing the vertical thrust component. This eventually gave way to the final solution, a pair of ceramic pincers which block the thrust of the motors as required.
Another interesting component is the landing legs. Made from light carbon fiber rods, they are released by melting a rubber band with nichrome wire and fold into place under spring tension. They also had to be carefully refined to absorb as much impact as possible without bouncing, which killed a few previous landing attempts.
Scrolling back through [Joe]’s videos and seeing the progress in his engineering is absolutely inspiring, and we look forward to his future plans. These include a functional scale model of the belly-flopping starship, a mysterious “meat rocket”, and the big one, a space shot to exceed 100 km altitude.
We’ve become so used to seeing SpaceX boosters land themselves back on the pad with clockwork reliability, that it’s easy to forget it took them a good number of attempts to get right. Inspired by SpaceX’s work, [Joe Barnard] of [BPS.Space] started working to replicate it at the model scale five years ago, with no engineering education or experience. On the latest attempt with a brand-new thrust vectoring Scout E rocket, he has gotten tantalizingly close to doing a controlled propulsive landing with a solid-fuel rocket motor.
We’ve all been thrilled to see the SpaceX rockets return to earth, landing elegantly on a floating pad. But those are liquid-fueled. The trick with a solid-fuel rocket motor is it can’t be throttled directly, which is a challenge when you need precision control to land. Thanks to [Joe]’s custom AVA flight computer and the remarkably consistent thrust curve of the Estes F15 black powder motors he used, it becomes a matter of igniting the descent motor at the right moment to make the vertical velocity zero at touchdown. However, [Joe] found that the time between sending the ignition signal and when peak thrust is reached was inconsistent, so he had to work around that. He did this by controlling how much of the thrust is spent in the vertical direction, by vectoring the motor side to side to spend some trust horizontally.
In this attempt, the rocket tipped over on landing due to excessive horizontal movement at touchdown. Joe tracked the cause down to a weak GPS signal caused by antenna position and a possible bug in the Kalman filter that fuses all the sensor data for position and velocity estimation. Thanks to incredibly detailed telemetry and logging done by the flight computer, data from every launch are used for future improvements. We are looking forward to the next flight in a few weeks, during which [Joe] plans to tune and test the control software, among other minor improvements.
Almost every single part of this rocket is a display of engineering ingenuity. The landing struts are designed to absorb as much impact as possible without bouncing while being light and quick to deploy. The ascent motor is ejected simply by moving the thrust vectoring mount to one of its extremes, allowing the descent motor to drop into place. The rocket also features a complete emergency abort system with a parachute, which can be activated manually, or by the flight computer if it calculates that landing isn’t feasible. We already covered [Joe]’s latest launch pad, which is a very interesting project all by itself.
When you’re building and launching a variety of advanced model rockets like [Joe Barnard], you don’t want to spend time building (and debugging) specialized flight computers for every rocket configuration. This challenge has led him to create AVA (All Vehicle Avionics), an impressive model rocket flight computer that he intends to use on all his future rockets.
All of [Joe]’s rockets feature active stabilization and guidance, and comprehensive telemetry using a variety of sensors. On the board there are three separate microcontrollers connected over I2C or SPI, each with its own micro USB port. The two smaller microcontrollers are both ATSAMD21s, also used on the Arduino Zero. The first is used for GPS and inertial navigation, and uses data from onboard and external sensors like the two IMUs (one is a backup), GPS and barometer to estimate the rocket’s position, velocity and attitude, The second is for telemetry, and it handles all external communications via a Bluetooth modem or long range 900 Mhz radio. The main processor (MPU) is a NXP MK20DX256 (also used on the Teensy 3.2), which receives data from the other microcontrollers and handles all the real-time operations and control outputs.
[Joe] gives a very detailed overview on the board, it’s capabilities, and the reasoning behind some of his design choices in the video after the break. Most of the sensors and microcontrollers were selected partly because of his experience with them. All three microcontrollers have Arduino bootloaders, also due to familiarity with the framework. AVA is the 12th in the line of flight computers [Joe] has built, and it is clear that a lot of work and hard-earned experience went into the design. Continue reading “Advanced Model Rocket Flight Computer Reaching For The Stars”→
While rockets launched from silos are generally weapons of war, [Joe Barnard] of [BPS.Space] thought model rocketry could still do with a little more thoomp. So he built a functional tube launched model rocket.
Like [Joe]’s other rockets, it features a servo-actuated thrust vectoring system instead of fins for stabilization. The launcher consists of a 98 mm cardboard tube, with a pneumatic piston inside to eject the rocket out of the tube before it ignites its engine in mid-air. When everything works right, the rocket can be seen hanging motionlessly in the air for a split second before the motor kicks in.
The launcher also features a servo controlled hatch, which opens before the rocket is ejected and then closes as soon as the rocket is clear to protect the tube. The rocket itself is recovered using a parachute, and for giggles he added a tiny Tesla Roadster with its own parachute.
Projects as complex as this rarely work on the first attempt, and Thoomp was no exception. Getting the Signal flight computer to ignite the rocket motors at the correct instant proved challenging, and required some tuning on how the accelerometer inputs were used to recognize a launch event. The flight computer is also a very capable data logger, so every launch attempt, failed or successful, became a learning opportunity. Check out the second video after the break for a fascinating look at how all this data was analyzed.
[Joe]’s willingness to fail quickly and repeatedly as part of the learning process is a true display of the hacker spirit. We’ll definitely be keeping a close eye on his work.
[Tweepy] flies unpowered aircraft, and he’d like to use the XCSoar flight computer app for gliders, sailplanes, and paragliders, but couldn’t find any hardware. XCSoar is an amazing app that can keep track of terrain, route, thermals, and a whole bunch of other variables that make flying more enjoyable, but running it on a device useful for a hang glider pilot is a challenge.
He eventually found a nearly perfect device in the Kobo mini e-reader. It’s e-ink, so it’s sunlight readable, uses a glove-compatible resistive touchscreen, runs Android, and is dirt cheap. The only thing lacking was a GPS receiver. What was [Tweepy] to do? Mod an e-reader, of course.
The electronic portion of the mod was simple enough; serial GPS units can be found just about everywhere, and the Kobo has a serial headers on the board. The case, however, required a bit of thingiverseing, and the completed case mod looks fairly professional.
With a few software updates, new maps, and of course the phenomenal XCSoar app, [Tweepy] had an awesome flight computer for under 100 Euro. The only thing missing is an integrated variometer, but a Game Boy will work in a pinch.
Since 2007, [Adam Kemp] has been leading a team of students from Thomas Jefferson High School, guiding them through the process of designing and building a small satellite that NASA selected for launch early next year.
The CubeSat, officially named TJ³Sat, uses commercial, off-the-shelf components for nearly all its systems. The team ran into a problem interfacing the FM430 Flight Module (PDF warning), so [Adam] designed an Arduino-based replacement. Based on an ATMEGA328, the entire board is a drop-in replacement for the FM430 Flight Module. On July 1st, the TJ³Sat will begin testing at Orbital Sciences Corp. to make sure the entire satellite is up to snuff.
The TJ³Sat’s payload will take data from the ground controllers and using a TextSpeak module convert serial data into spoken voice. This audio will then be transmitted over amateur radio frequencies and will be picked up by hams all over the world. We’d like to wish the students at Thomas Jefferson High a hearty congratulations for being the first High School to build a satellite and hope the testing and launch go as planned.