wind_sensor:meeting_minutes_apr_28_2017

Attended: Mengyuan, Creighton, Scott

  • Updates:
    • Inaccuracies in single microphone wind speed algorithm due to assumption that all microphones are set the same way
    • To fix this, microphones were re-calibrated individually, each microphone calibrated with 8 wind speeds
    • Peak detector coefficients were determined for each microphone
    • Raw wind direction tests were conducted from 0 to 360 degrees using the new peak detector equations
    • Peak detector outputs of each microphone with the north microphone at 0 to 360 degrees were plotted
    • Using this raw data, microphone angle and ratio of the two largest microphone readings at each angle were plotted
    • A linear relationship was derived mapping the microphone angle to this raio
  • To Do:
    • Conduct tests to verify relationship
    • Look at implementing some other filters (i.e. FIR)
  • Updates:
    • To try and fix the problem we were having last week (zero-crossing detector not transitioning), we added a DC offset to our receiver output as well as the zero-crossing detector. This DC offset put the received signal within the operating voltages of the LM339 comparator chip we were using. This prompted our circuit to start working again and we can see the results below (the yellow is the emitter and the blue is the output of the ZCD, letting us know when the receiver heard the pulses.
    • After that was fixed, we redid our perfboard to a more “permanent” setup (we just stripped wires and made everything flat). Of course we tested it to make sure it was still working, which it was.
    • We recently found the option to overclock the Teensy to 120MHz, so now our Teensy runs at 120MHz. It was good that the code didn't have to change much.
    • Finished the program to measure propagation time. Essentially it worked as follows:
      • We send 10 pulses and record 10 response times. We calculate the difference by taking the difference between the first response and the first emit time (we chose not to average all 10 times because they were somewhat inconsistent–though we should revisit this later on). We then wait approx. 1.5us for the “ringing” to die down.
      • We do this 500 times and average all the differences. This produces our propagation time.
    • When testing our program with our setup, our results were pretty consistent (difference += 0.4us). We decided to start taking data points and testing with the anemometer.
      • The first thing we did was plot the measured propagation times at different wind speeds.
      • We measured the propagation time with no wind speed and with the three speed settings on the black fan (note: we also tested it with the grey fan but because we may have shifted the box during the movement of the fans, the results weren't as consistent. We chose to omit them, but we will have to revisit this)
        • These results are in line with our initial assumptions: increasing wind speeds in the direction of propagation will reduce the propagation time.
        • There was one issue however: at the distance that we were testing, we expected our propagation time to be ~240us. However, our measured propagation time was around ~340us, which is significantly different. We believe this may be due to some constant time delay added into our measurements.
      • To calculate this constant time delay, we use this equation:
        • t_meas is our measured propagation time from our setup, V_wind is the measured wind speed using the anemometer, and V_sound is the speed of sound. We plotted t_meas against the inverse (V_wind + V_sound) to get this plot:
        • Looking back at the equation, we see that the slope of this plot represents the 'd' or distance, and the y-intercept of this plot is our time delay. Now that we know all our parameters, we can calculate the wind speed using the following equation:
    • We also wanted to see what would happen if we plotted the wind speed against the measured propagation time. We were surprised to see that it was also very linear:
      • This meant that we could go directly from our measured propagation time to the wind speed, with the only caveat being that it would only work for a given temperature (since the speed of sound changes with temperature)
    • In other news, we also finished our poster for the poster presentation and the final presentation for the SCEL general meeting.
  • To Do:
    • Instead of just using the first set of times to calculate the difference, try use all ten sets to see if the overall average becomes more stable.
    • Add in calculation of wind speed into program so it can print out the calculated wind speed directly, and conduct tests with more wind speeds and the other fan.
    • Experiment with amplifiers to see if we can increase the SNR (signal-to-noise ratio). This should give us better results. Also, see if we can utilize use the Schmitt trigger's two threshold levels to have one act as a “qualification” and the other as the actual zero-crossing detector.
    • See if we can reduce the ringing of the emitter by sending 180-degree phase shifted pulses after we send our actual pulses (destructive interference).

Authors

Contributing authors:

mwu snishihara

Created by mwu on 2017/04/30 10:38.

  • wind_sensor/meeting_minutes_apr_28_2017.txt
  • Last modified: 2017/05/13 20:57
  • by snishihara