Date: 5-6 May 2007
Report
The aim of the Kestrel project was to design and construct a cold-gas actively stabilised model rocket. I had been working on the idea since July 2006, and it had been very satisfying to see the rocket actually take shape over the final couple of months (although it had involved a couple of late nights towards the end of the week to get the rocket launch-ready – many thanks to the Cambridge University Spaceflight brigade for keeping me well supplied with tea, biscuits and pizza!). UKRA Safety and Technical Committee approval had been obtained for the launch; nonetheless, I had extreme doubts as to whether the control system would work at all...
We arrived at the East Anglia Rocketry Society's Elsworth launch site mid-morning on Saturday, to find quite a welcoming committee waiting to see my rocket (mostly my fellow Cambridge engineers). I set up in the workshop tent to implement the last part of the software architecture; however, the number of interested spectators made it quite hard to concentrate on coding! With the launch sequencer implemented and batteries and gas cylinder installed, I tried a “spike test” - running the attitude control system with the vehicle balanced on a pointed metal rod. The result was severe instability at the beginning of the “flight”, but little evidence of control activity towards the end. I concluded that the pressure regulator was set to too high a pressure (something that had not been obvious in tests in the lab with the 5 bar compressed air supply) – however another test, with the regulator unscrewed a few turns, showed control action that was too weak, again to the disappointment of the expectant crowds.
At this point Damian stepped in and kindly offered me the use of his caravan's awning – with tables, chairs, 240V for my laptop (whose battery by now was flat) and, most importantly, a modicum of peace and quiet. With help from Brendan and Octave (and tea and sustenance from Damian and my father) I designed and implemented a rather more scientific control law (a 2-axis lead controller) to replace the tweak-requiring PID loop I had previously been using. After some experiments to determine the nozzle thrust for different pressure regulator settings, I ran a few further spike tests with rather disappointing results. However, I explained these by saying that the dynamics of balancing on a spike (as either an inverted or a non-inverted pendulum) were not the double-integrator dynamics of a rocket in flight, and thus decided (with some encouragement from others) to run a test flight in the last light of the day with the pressure regulator screwed further in than in the spike tests to ensure high gain.
A crowd duly gathered as I and others prepared the vehicle for launch, using the spike's aluminium base plate as a flat launch pad that would hold the launch detect switches closed until liftoff. The E9-P motor was installed (after a little bodging to get it to fit in the motor mount!) and connected, and the electronics was armed. However, as I armed the pneumatics (by pulling out an allen key and allowing a valve to close) I heard a loud hiss of escaping gas; within a few seconds the cylinder was empty (it seemed gas was leaking through solenoid valve number 3), and the launch attempt had to be scrubbed for the day.
The next morning was windy and cloudy, so few high-power rocketeers were intending to fly. However, Kestrel with its 1.5m projected apogee had no such restrictions (as long as a sheltered launch pad could be found) so I set about retesting the control system. I speculated that the previous day's leak was due to exceeding the valves' pressure ratings (the cylinders contain liquid CO2 with a vapour pressure of 50bar at room temperature and the valves were only rated to 8 bar, hence the need for the regulator), so my father and I ran tests to calibrate the no-flow pressure of the regulator, and the thrust produced with a solenoid valve's pressure drop in circuit, as a function of the screw position. It did indeed look as if overpressure had caused the previous night's leak; armed with the correct screw setting, I set about reassembling Kestrel for testing. Unfortunately I managed to crack a plastic valve body through over-enthusiastic teflon-taping of a blanking plug, but reduced-viscosity superglue saved the day. A test on the spike indicated appropriate control action (an attempt to stabilise itself at least, though the zero point was clearly drifting) and I decided to go ahead and try for a launch.
Once again, the aluminium base plate was laid out in a sheltered spot behind the carpark; once again, an E9-P motor was installed and all systems made ready. The three videoers, numerous photographers and the two armers (myself for the avionics and pneumatics, another EARS member for the launch controller) took our positions and negotiated our escape routes. Then, in a few seconds, the vehicle was readied, beeped to indicate it was ready to launch, we ran away, and the motor lit...
...and the ungainly rocket lifted slowly half a metre off the ground, turned over and plummeted headfirst into a thistle in a cloud of smoke. The motor burnt for a second or two longer, then I was running towards it to retrieve my construction. I could hear a hiss of escaping gas, then an intermittent fizz, which brought me up short: what if the lithium-polymer batteries had short-circuited (despite the fuse I had installed) and were, in the words of one EARS member, “about to pull their self-destruct trick”? My father ran off to retrieve a screwdriver to open the airframe and extract the battery packs, while the rest of us stood around at a safe distance; when after a few minutes there was still no smoke or other indication of a conflagration, I breathed a sigh of relief and took a closer look at the damage.
Because I had assumed it would land on its feet, the lower section of the airframe was aluminium while the electronics were housed only in a Sainsbury's Basics plastic box; the nose-dive had severely damaged this, and on closer inspection the collision had also forcibly desoldered some surface-mount components from the processor board and removed one of the gyros. A search failed to locate the missing £50 gyro, although some more fragments of plastic box were recovered. Then it was time to pack up and depart, to get my father to the station to catch a train.
Overall, it was a disappointing first flight for Kestrel – not least because I had told many people I intended to fly on Saturday morning and did not end up launching until Sunday afternoon, for which I apologise. The behaviour of the rocket in flight appeared to experienced observers as if there were no stabilisation whatsoever; that and the fact that the CO2 cylinder seemed almost full after landing suggested that the control system did not after all engage (although the avionics were beeping and clearly waiting for launch-detection before the motor ignited). Unfortunately I had not implemented datalogging in the flight control software, though all the hardware was in place to support it, so I do not know what (if anything) the control system thought it was doing in flight; this will be the highest-priority modification for the next version. Fortunately because of its very low apogee Kestrel requires no special dispensation to fly – thus I intend further tests over the coming weeks, with a view to having a working vehicle for my project report and for the next EARS launch day. If at first you don't succeed...
Coda: in the rush to pack up I had left my lab notebook in the workshop tent, where it was found by a member of the EARS committee. He was kind enough to return it to me a couple of days later, along with a prize I had apparently won in the event's raffle: an Estes ready-to-fly “NSA Starship” starter set. I hope that rocket does not prove to be the only one I can fly successfully!
Lessons learnt
Rocket redesign issues
Items I took that were (unexpectedly) useful
Items I didn't have and (could have) needed
Items I took that were (definitely) unnecessary