Testing the BoxDAQ

The BoxDAQ project has been a whirlwind of excitement so far. It started in December of 2016, when I bought a VectorNav VN-200 Rugged INS to use for improving the accuracy of our air data calibration maneuvers. I was really impressed with the data, but it wasn't long before Wasabi called me up and said they had an Extra 300L with an empty front seat that was going to be spending some time upside down in the near future. I'd written some software to log data from the INS onto my laptop, just like we've done for air data, but this proved to be kind of clunky in the context of aerobatic flying. We put the laptop in a backpack that I wore front-side, and we swapped out the Softie backpack chute for my Strong seatpack to scooch me far enough back to get full stick travel.

The flight went great and we got really awesome data. Knowing that the INS could handle a 360 deg/s roll rate and sustained asymmetric G, this system needed to be boxed up into something a bit more usable. I put the 3D printer to work and whipped out the soldering iron, and before too long, the prototype BoxDAQ saw its first flight in my KR2S.

At this point, the project presented a bit of a challenge. Flight testing an air data boom or a trailing cone, while time intensive, is a problem that statistic knows how to handle. You take some data, compare it to a truth source like a four-leg GPS calibration or an INS, put error bars on both sides, and you have a pretty good feel for how well the system is performing. An INS is a bit of a different animal - in the case of a commercial-grade unit like the VectorNav, it'll tell you when its solution is out of spec, so the test becomes more of an evaluation of what types of maneuvers will make the INS's internal filtering unhappy.

I jumped at the chance to fly the BoxDAQ on the Extra again. Wasabi had a customer who wanted to do spin testing to evaluate the SARO rating system, which meant the BoxDAQ was going to see quite a variety of spins over the course of two sorties.

I was fortunate enough to FTE for the second sortie, and we were able to include some test points specific to the BoxDAQ. One of the potential markets for the system is a Red Bull Air Race team, so we included a 6-G slalom course followed by a vertical reversal. We also looked at roll captures and left and right snap rolls. Combined with upright and inverted spins with a variety of aggravation, this is a pretty good shakedown for the INS and a good test of the toughness of the hardware.

I got home with a sore neck, but dove into the data the next morning. As usual, the tricky part is deciding how to look at the data. The spin plots were inspired by a plot of yaw rate vs. number of turns that shows up in Kershner's Basic Aerobatic Manual, and are really, really cool if you're into that kind of thing.

From a data validity standpoint, we can see that the INS drops from status 2 to status 1 in many of the spins. This means that the INS has sufficient dynamic motion to generate a solution, but it doesn't meet the accuracy specs provided by the manufacturer. The working theory is that this might be related to blanking of the GPS antenna as the airplane tucks through the inverted incipient phase. The data looks good in these portions of the spin, but we can't quite say how good.

I was really, really happy with the data from the slaloms and the snap rolls. There were no status drops, and the INS was perfectly happy through a G onset from 1 to 8Gs in half a second at a roll rate of 360 deg/s. The slaloms presented repeated rapid-onset G in varying attitudes followed by several seconds of sustained high-G and likely GPS blanking in the vertical reversal without a loss of solution status. This gives me a ton of confidence in the INS and its ability to handle very challenging flight environments.

I have a little bit more analysis to do on this data, but overall I'm really happy with it. I'm confident in the BoxDAQ hardware and I'm looking forward to getting the final rev out the door in May of this year (the unit is available for pre-order). I really can't wait to dive into some inertial alpha reductions and some other cool visualization stuff as well.

Big thanks to Elliot and Tyler for looping me in on their Extra shenanigans. It's a big opportunity to fly hardware on an airplane like the Extra when it's early in the development lifecycle. Check out the brief and video of more of their testing here!