We are having some odd intermittent problems on our fleet of RVJet/Pixhawks so I took a good look at the +5V servo power rail with my scope to see if that was the problem. I was quite surprised to see power spikes as high at 400mV when the servos were thrown. The ESC also puts some noise on the rail of about the same magnitude, even if the internal BEC is not used. Here's a shot of the scope from one spike (100mV vertical division, 200us horizontal).
Some of the planes had a separate high quality Castle Creations BEC and some used the BEC built in to the speed controller. Some planes had high quality servos (HiTec or Airtronics) and some had Range Video house brand. No matter what hardware configuration the noise levels were about the same.
I put a 680uF electrolytic cap across the +5/GND of the servo power rail and the noise levels dropped dramatically. The noise from the servos was now down to less than 100 mV and the motor put slightly more noise on the rail. These noise spikes were very infrequent; I had the scope on single shot mode and often it took about 30 seconds of constant servo movement to generate one.
So I have a few questions:
1. Is there a consensus on how much noise is acceptable on the servo rail power bus?
2. Is a large bypass cap the best way to deal with it? I don't think a zener diode would work this close to the setpoint voltage.
3. Could the 400mV noise spikes been the source of odd, but very intermittent, behavior from the Pixhawk? For example, we've had a couple of our planes porpoise wildly on occasioin. Sometimes our airspeed indicator rapidly jumps to very high settings when the plane is on the ground (e.g. 12 m/s). And we find we have to do frequent power-down reboots to get the system working properly.
Thanks in advance for any tips!