It doesn’t fly.
I finished fabricating the second blade at around midnight. After that, it was time to balance them. With some slightly crude instruments, I got the blades to within about 0.1 – 0.2g of each other, and the difference in their length-wise center of gravity was within about 1-2 millimeters. I thought this should be good enough to get the heli decently stable with the blades spinning. A slow spin-up instantly revealed that it was NOT good enough. The thing shook so violently that I thought for sure it would rip itself apart, or at LEAST bend some of its shafts if I tried spinning up to flying speed.
The vibrations weren’t the only thing to break my heart. Before spinning up the blades, I sent them each a short program that just listened for the hall sensor input and toggled all of the lights when it was seen. So after every revolution, the lights should switch between being on and being off. This worked fine when I passed the magnet in front of the hall sensor manually, but when I started spinning the blades up, it seemed that the input was missed the majority of the time. The lights would sporadically react after any number of revolutions. After an hour or so of trying to add bigger magnets, more magnets, and moving the magnets closer to the sensor, all to no avail, I finally decided to take a close look at my sensor’s datasheet.
I originally chose the AH182 hall effect latch because it used very little power to operate. As it turns out, the way the AH182 achieves such a low power requirement is by sampling at a rate of 20Hz with a duty cycle of 0.1%. This means that 99.9% of the time, the sensor is not even on! Unless the sensor happened to check for the magnet at the exact moment that it passed over it, it would completely miss the revolution. In practice, this meant that any faster than about one revolution per second and the blades simply wouldn’t work. I needed a sensor that operated continuously. The replacement I chose was the Honeywell SS451A. A careful, thorough read through the datasheet did not reveal any issues with this device.
I replaced the old sensors with the new ones when they arrived, and they worked perfectly. To remove the old sensors from the aluminum to which they were CA glued, I submerged the glued joint in a small dish of acetone and heated it to ~80 °C. After sitting in there for 5 minutes or so, I was able to pry the two apart pretty easily and clean up the excess residue on the aluminum.
As for the vibrations… I thought the length-wise center of gravity and mass were balanced well enough, but perhaps the z-direction misbalance was causing the problem. I don’t know if I’ve mentioned it yet, but all of the heavy electronics were mounted to the top of one blade and to the bottom of the other. The electronics covered 2-3 inches of the blade, and I didn’t want that portion to be without lights, so the lights were mounted on the opposite side of the electronics. But the lights had to be mounted on opposite sides of each blade so that they’d be visible from both the top and bottom. As a result, the electronics were mounted on opposite sides of each blade too. Now that I had this awful misbalance, I regretted this design decision and wished I had just mounted the electronics symmetrically. I could live with less lit-up area if it meant the thing would fly!
To try to remedy the z-direction misbalance, I removed the battery on each blade and mounted them on the opposite side of the circuit board. This would not give me perfect z-balance, but it would certainly be much closer than before. Again I did a semi-decent job of balancing the weights and length-wise CG of the blades, and again I had terrible vibrations upon spinning them up. It seemed the z-balance may not have been the main problem here. It was time to turn to the experts.