James Dougherty wrote: > Regarding the firmware and how it can fail with harmonic oscillations > and sensor data, the "Nyquist rate" is the rate in digital signal > processing which is the upper bound for how much information you can > process on a bandwidth limited channel. The nyquist rate is 2x the > true data rate. Nyquist says that if you want 100 samples/second you > should actually Be sampling 200 samples/second. > We kind of know that intuitively when we fly altimeters and crank > up the sample rate as it gives us better graphs As to how the flight > computers all operate with sensor data (Barometers, Accelerometers, > IMU's, etc) a rough example of what could be happening is that if > the A/D samples 50hz, and you oscillate at 20hz, your effective > sampling rate is 1/2 50hz (25hz) and half the Nyquist rate is 12.5hz, > the firmware (as it loses efficiency in sensor data) effectively > "slows down" and apogee may never be seen! Yeah, half the sample rate is the highest possible frequency that can be represented in the data; anything higher than that will come through at the difference between its frequency and the sample rate. (So at 200 Hz a 190 Hz signal will appear as 10 Hz in the output, and a 200 Hz input, exactly matching the sample rate, can appear as an arbitrary DC bias depending on how the sample timing lines up with the phase of the incoming signal!) > This is an evil side effect of vibrations with noisy > sensors driven over A/D Yeah, this is what we ran into. I think the AltAcc sampled at something like 16 Hz and I forget what the rate was on the G-Wiz, but one way or the other we clearly were seeing vibrations with a lot of high-frequency content well above the "Nyquist" rate of either system. The acceleration trace downloaded from the flights with successful timer or baro deployment - from an R-DAS recording at 200 Hz - was full of artifacts that looked like "beat frequencies" between the sampling frequency and higher-frequency stuff in the signal. When I saw that, and compared it with output from the same unit on solid- propellant flights, I thought "if I get this mess when sampling at 200 Hz, then no wonder the acceleration-based software on the other devices, using lower rates, got confused". > When will you fly your rocket again? Well, this was back in 2002 as I mentioned; that configuration is pretty much retired; the next one will be smaller, I think. (KISS was about 7-9 feet long and 6 inches in diameter, and designed more for simplicity and robustness than performance.) > Do you come to TCC launches? Not as much as I used to, but I still make it out there every once in a while. -dw