Built on the knowledge and lessons learned through the Scout program, the Echo series of vehicles are designed to be scalable for incremental testing. Trial flights for the Echo series began on October 27th, 2016, and ended on December 21st, 2016.
Designed with a similar recovery system to Scout, Echo carries all the needed hardware and software to land vertically under retro-propulsive thrust. Beginning with a series of low altitude stability and recovery tests under parachutes, Echo will only scale up to higher altitudes and higher-stakes landing attempts after the technology has been tested and proven in a low-risk environment, making the program much less costly than the full-stack style tests of the Scout series.
The primary focus of the first Echo Test Vehicle was to assess the tuning of the PID thrust vectoring algorithm, as well as the accuracy of flight simulations created in MATLAB and Simulink. Scout V 0.8 - V 1.0 used a PID stabilization algorithm, but never flew reliably enough to obtain accurate data about the system in flight. Previous Scout vehicles also had issues with apogee predictions based on vehicle weight and a fairly stable flight profile. Though no vehicle ever flew stably, it was clear there were errors in the simulation.
With the errors fixed and the PID system tuned to what seemed accurate on the ground, Echo TV1 flew just a few degrees off axis to just under the flight simulation's predicted apogee of 30ft. The test proved that the PID system needed only slight tuning to achieve reliable stability, and the flight simulation's errors had been fixed. The only part of the test that failed was parachute deployment, due to a bit of unreliable code triggering the chute deploy relay onboard.
The Echo Test Vehicle 2 served to further tune the PID stabilization algorithm, as well as test in-flight deployment of the landing legs. Both tests returned nominal results - a nearly flawless ascent to 30ft, and lockout of all four landing legs after being deployed at 20ft AGL. Once again, the parachutes failed to deploy in time to ensure a soft landing on the ground. Though they deployed shortly after the landing legs at just under 20ft AGL, there was not nearly enough time or airspeed to inflate both chutes before the rocket slammed into the ground. One chute did inflate, but only a few feet from the ground, which proved fairly ineffective.
With stability improved, and more height to deploy parachutes needed, Echo TV3 would be modified to ascend to approximately 70-100ft AGL, with similar parachute and leg tests during the flight.
Echo TV3's main purpose was to fly twice as high as Echo TV1 and TV2. This would theoretically allow for the parachutes to safely release, and more realistically simulate the flight profile required for propulsive landing. The rocket lifted off quickly as expected, but slowly pitched over to one side. By the time of main burnout, Echo was completely horizontal. Both chutes were deployed just after apogee, but did not inflate at all, and the rocket slammed into the ground ripping the airframe in half.
The cause of the pitching over of the rocket was due to the inaccuracy of the Complimentary Filter used to measure the physical orientation of the rocket. The filter looks for the gravity vector among all 3 axis of an onboard accelerometer, which is greatly skewed under thrust. This led to the rocket assuming it was flying straight, only correcting for movement accumulated by the gyroscope. The flight was unintentionally a demonstration of an entirely gyroscope-based guidance system.
With new code to test, Echo TV4 pitched over more quickly and more violently than Echo TV3, but this time, due to a mechanical error. The chutes deployed successfully for the first time, as well as the legs, and provided a soft landing for the rocket.
Without much thought, the cable used to pass current through the leg release system was threaded through a small opening used to adjust the mechanical connections of the TVC mount. At the time of launching, the cable wasn't secured, and it ended up obstructing one axis of thrust vectoring mount. The onboard data recorded shows that while the vehicle pitched over, the obstructed axis tried harder and harder to correct, but was ultimately unsuccessful.
The new code onboard attempted to reduce the inaccuracy of the gravity vector measurement by accounting for the thrust curve and net force of the main booster. Unfortunately, given the nature of the failure, the code could not be verified, and most of the guidance data from the flight could not be considered as accurate.
Below is a rather silly comparison of Echo TV4's flight to an ILS Proton M's flight anomaly in 2013.
Flying in a similar fashion to Echo TV3, TV5 pitched over a bit as it flew. Though the onboard code had been modified to account for the skewed gravity vector, Echo TV5 revealed that the complimentary filter combining the gyroscope and accelerometer was fundamentally flawed. Originally chosen for its simplicity and ease of use, the complimentary filter used both sensors combined for orientation onboard.
After Echo TV5's flight, I decided to scrap the complimentary filter and current IMU altogether in favor of a new, more robust solution, described below.
For Echo TV6, a new Inertial Measurement Unit (IMU) was used. The addition of a magnetometer on the new IMU allowed for much more accurate measurements and sensor fusion. Filtering of all 3 sensors in the IMU (accelerometer, magnetometer, gyroscope), also took place onboard the IMU instead of the main flight computer. This allowed for a faster control loop, which can be used to explain a chute deployment anomaly during flight.
TV6 maintained accurate orientation measurements through the whole flight, exceeding the previous systems capabilities by a long shot. The vehicle experienced a bit of a wiggle on the way up - this was due to the now un-tuned PID loop. While previous flights had a more solid tuning, the orientation measurements used to calculate TVC correction angles were incorrect. The flight software was tuned with about the right amount of P-value without enough D-value.
Just before burnout of the main motor, Echo's chutes deployed. One of the safety features in the flight software is designed to deploy chutes in an anomaly scenario where the altimeter isn't functioning properly. The if-statement checks if the vehicle is 10ft AGL, and if a certain amount of time has passed since liftoff. The mistake was a porting error, as the timing contingency referenced the number of control loops executed, NOT the number of milliseconds passed. The burdened flight computer had a faster control loop during this flight, and deployed the chutes early because of this.
Flying only days after TV6, Echo TV7 served to further tune the PID stability algorithm. Another wiggly flight, but this time due to an excess of D-value. Had the sample rate been higher (1/100th of a second on this flight), this might not have been an issue. Unfortunately due to cold weather, many of the launch coverage camera batteries failed just before liftoff and the launch proceeded without them. Fortunately, at every launch, there are plenty of cameras for redundancies and cases like this - below are the best angles of the full flight.
The mission objective of Echo TV8 was to test the drag fin system used to passively stabilize the rocket on the way down before a retro-propulsive landing. The flight profile was intended to send Echo to roughly 130ft AGL, then release the drag fins for the first half of descent to observe their effect on the vehicle's orientation. At 65ft AGL the chutes were planned to pop out to ensure a soft landing on the ground. In keeping with the Echo series global objectives, no retro-propulsive landing can be attempted until all necessary systems have been tested and qualified in flight.
The chutes deployed almost immediately following apogee, along with the drag fins. Initial flight data showed the vehicle only reaching about 55ft AGL, so the flight software did its job correctly there. However, after tracing backwards and rebuilding the offline flight simulator, it was found that the only way the rocket could've reached just 55 ft was with less than 80% thrust. Multiple separate simulators predicted apogee of 130ft as expected with 100%(w/ ± 5% tolerance) thrust. This lead to a manual visual analysis of the flight from a far away range camera.
Using this camera, the known height of the Echo vehicle, and a bit of Photoshop to correct for lens/perspective distortion - apogee of Echo TV8's flight was found to be within a range of 115-125ft. This is likely quite close to an accurate measurement - the offline simulations run did not account for any drag.
After Echo TV8's flight, a number of altimeter based experiments were performed to find the root cause of the bad reading. The cause was determined to be within a poorly written/referenced library used to compile the flight software, and all hardware was observed to be working just fine.
The software was then modified to use a new library for the barometer/altimeter measurements, fixing the issue. A secondary issue from TV8's flight was that one of the drag fins failed to deploy, due to the cold weather's effect on the deployment rubber bands, easily remedied by adding more.
Echo TV9 was the first BPS launch to be webcast (live with a 4hr delay), which is unfortunate considering it was the worst flight of all the Echo vehicles.
TV9 lifted off, slammed into the launch tower, and spun out of control immediately. The vehicle recorded an ascent to about 30ft AGL, before falling back to the ground. The primary cause of this failure was poor hardware calibration in the TVC mount. The mount contains a set of actuation pushrods and connecters, secured with adjustable linkage stoppers. The preflight standard operating procedures require that these linkage stoppers be zero-ed out and re tightened before each launch to ensure a properly calibrated mount, however it's likely than in transit to the launch site, one of these slipped out of position slightly.
The vectoring mount design used here is tedious to assemble and disassemble, and accessing these pushrods and connectors is impossible on site. The uncalibrated mount went unnoticed, even through the on-pad TVC range check, which did NOT include any calibration tools, only visual inspection. The solution for this type of problem is two fold. First, a new vectoring mount with access panels for on-site hardware checks will be designed. Second, a custom TVC mount calibration tool will be used to check alignment before each flight from now on. This tool will double as protective cover during vehicle transit to and from the launch site, further protecting the mount from external interference before flights.
After Echo TV9's launch failure, the series was ended, in favor of pursuing development of a slightly larger and more powerful launch vehicle called Relay. More details about the next series of BPS rockets can be found by clicking below.