Characterization of Unmanned Aerial Vehicle Noise Kyle Hahn Graduate Program in Acoustics, the Pennsylvania State University College of Engineering Noise measurements of unmanned aerial vehicles (UAV) have not been extensively published. With personal and commercial UAV usage increasing significantly, it is necessary to provide accurate and controlled UAV noise measurements for future work. A line array was used to record a professional grade UAV in an anechoic chamber to obtain general sound pressure levels and frequency spectrums. It was determined that the UAV contained strong harmonic tendencies up to 5000Hz, and general noise characteristics from 5000Hz to 20000Hz with the amplitude decreasing as frequency increases. I. Introduction In the past decade, unmanned aerial vehicles have become extremely popular among private companies and the general public. According to Google Trends, the frequency of Google searches related to the term unmanned aerial vehicle have steadily increased from the early 2000s and peaked in popularity in December of 2017 (Google Trends, 2017). UAVs were developed by the military to execute missions too dangerous for soldiers to undertake. With new interest by the public, a need has arisen to regulate public use. UAVs can be easily equipped with high quality video recording and live video streaming capabilities and are almost limitless in maneuverability. As a result, UAVs pose a real threat to personal privacy. UAVs are loud, disruptive, and true to their alias, drone, emit a constant and irritating buzzing sound. If they are to become a part of everyday life, by usage in public areas and for commercial package deliveries in populous regions, the effect of their noise output must be studied and characterized. II. Experimental work The UAV used for this study was a DJI Phantom 3 Quadcopter. The UAV was flown in the acoustic anechoic chamber, located in the Garfield Thomas Water Tunnel uilding on the main campus of the Pennsylvania State University. The chamber s working volume is approximately 9.3m high, 5.5m wide, and 6.7m deep. The walls and floor have fiberglass wedges measuring approximately 0.914m deep. This chamber meets the requirements of ISO 3745 (ANSI S12.55) and IEC 268 from 80Hz to 12.5 khz. The chamber was configured to be a full anechoic chamber with wedges placed on the floor. To record the audio data, nine one half inch PC-ICP microphones with wind screens were arranged in a line array. The microphones were spaced in 50cm increments to utilize the full width of the chamber. Fig 1. Microphone array with 50cm spacing, showing the two positions at which measurements were taken.
A sample rate of 48000 samples per second was used. The microphones were raised to a height of 1.19m above the floor, and a height of 0.356m above the highest reach of the floor wedges. The UAV hovered at four locations with two orientations. The first locatation was directly over the center microphone at position one, hovering at approximately 3.35m measured above microphone five. The second location was directly over the end microphone at position two, hovering at approximately 3.35m measured above microphone nine. Measurements were taken additionally over positions one and two at approximately 6.71m above microphone five and nine respectively. The farfield of the UAV s sound radiation was assumed to be at five UAV diameters, which corresponds to the hovering heights of 3.35m. Additionally measurements were taken at the height of 6.71m in both positions. One diameter of the UAV was measured as the distance from the tip of one propeller to the tip of the diagonally opposite propeller. Two orientations were recorded at each position and at each height. The UAV was oriented at 90 degrees, with the front of the UAV perpendicular to the microphone array, and at 45 degrees with the front of the UAV oriented at 45 degrees relative to the microphone array. The air temperature during the measurement was 25 degrees Celsius, with the humidity at 50%. Due to unavoidable circumstances, the lights remained on in the anechoic chamber and three persons were present during the recording of the UAV. Each recording was between three to four seconds long while the UAV was hovering and stationary. While the term stationary is used, in reality the UAV was very difficult to keep in a single, stationary position and would move around the correct position with the pilot making manual adjustments. The maximum displacement due to drift was approximately 0.5m in any direction. III. Results and Discussion Figure 2A and 2 show the power-spectral density (PSD) of all four locations with both d re Pa 2 /Hz d re Pa 2 /Hz 90 degree and 45 degree orientations. All eight of the measurements show good agreement on the locations of the first seven peaks, or up to 1500Hz. After 1500Hz, the peaks widths increasingly spread out until peaks are indistinguishable from noise. The noise floor from an ambient noise measurment was at -200 d re Pa 2 /Hz at 80Hz, and drops to -270 d re Pa 2 /Hz past 5000Hz. The UAV signal is well above the noise floor, even past the 20kHz range. A Figure 2 Frequency (Hz) Frequency (Hz) Figure 2A shows the PSD from 80Hz to 24000Hz. Fig 2 shows the peaks up to 5000Hz. The f of the PSD is 0.7324 Hz. Each of the UAV s four propellers are rotating at nearly the same frequency. The slight differences in the fundamental frequencies are exacerbated in the upper harmonics, causing the increase in the spread of the harmonic peaks. The first 10 peaks (including the less prominent peaks) are shown in Table 1.
Peak Frequency d re Pa 2 /Hz 90.28-90.97 180.2-72.96 271.7-94.74 362.5-75.6 451.2-100.9 542.7-77.31 631.3-102.2 723.6-74.06 810-95.98 904.5-81.18 Table 1 The first peak is located at 90Hz with each consecutive peak being a multiple of the fundamental. The even numbered harmonics are consistently 15 to 20 d lower than the odd numbered harmonic peaks. One possibility for the nature of this behavior lies in the orientation of the propeller blades. On a UAV with four symmetric propeller locations, it is necessary to have the net torque created by the rotating propellers to be zero. This is accomplished by the front right and back left propellers rotation counterclockwise, and the front left and back right propellers rotation clockwise. If the propellers were synchronized to produce the same phase, the UAV could be modeled as a lateral quadrapole. ecause the four propellers are free to move completely independent of each other, a lateral quadrapole is unlikely. Nonetheless, this orientation could be responsible for amplifying the odd numbered harmonics. The theoretical fundamental frequency of one propeller was calculated using the following equations: F = ρ πd!!.! d! fp 4! kp! Eq. 1.1 from the comparison of 149 real static thrust data points to the theoretical thrust (Staples, 2014). Since the UAV is stationary, the thrust must equal the weight of the UAV. Solving for f, equation 2 results: Eq. 1.2 f = 4Fk!.! ρπd!.! p! Next, multiply f by the number of blades per propeller, n b, divided by the number of propellers on the UAV, n p. Eq. 2 f! = f n! n! The theoretical value of f! was calculated to be 88.84Hz. Figure 2A shows the first peak to be 90.28Hz, which is within 1% of the theoretical value. Where F is the thrust, ρ is the air density, d is the propeller diameter, p t is the pitch, and f is the rotations per second of one propeller blade (Staples, 2014). The pitch of a propeller blade is defined as the distance the propeller would travel through a soft solid, like a screw into wood. In calculation, ρ was 1.255 kg/m 3, d was 0.24m, k was equal to the correction factor of 3.295, the pitch was 0.127m, and F was calculated to be 12.557N. The correction factor is purely empirical and corrects for static thrust estimates. It was determined
A D C Fig. 3 showing the reciprocity measurements of position one with the following orientations: 45 degrees at 6.7m, 90 degrees at 6.7m, 90 degrees at 3.35m, and 45 degrees at 3.35m respectively The dashed line shows channels 1-4 in the microphone array, and the solid line shows channels 6-9. The error in this measurement was produced from the uncertainty in position of the UAV from drift. Fig. 3C and 3D show good reciprocity, while Fig. 3A and 3 do not show strong reciprocity. In the chamber, a large amount of wind turbulence was observed coming from UAV, producing wind noise. It is possible that at the lower height, the wind did not reach the outer microphones and so did not interfere with the sound measurement of the outer microphones. However at the higher height, the turbulence was able to spread to all microphones in the array, causing discrepancies in the reciprocity. This turbulence also caused discrepancies in the expected sound pressure level differences between the higher and lower measurements when the UAV was in the same position and orientation. Figure 4 shows these comparisons.
A D C Fig.4 showing the SPL comparisons between the following: 45 degree orientation in position 2, 90 degree orientation in position 2, 45 degree orientation in position 1, and 90 degree orientation in position 1. The dashed line shows the sound pressure levels of each microphone channel for the higher position at 6.71m, and the solid line shows the d sound pressure levels at the 3.35m height. In general, each microphone channel recorded higher SPLs when the UAV was in the 3.35m position when compared to the SPLs of the 6.7m location. Some anomalies exist, such as channel 2 from Fig. 4C which recorded a higher SPL when the UAV was in the 6.7m position. The most likely explanation for this discrepancy is due to the unpredictable wind noise.
A C Fig 5 shows the SPL levels compared at the same heights with the 90 degree and 45 degree orientations. Fig. 5A shows the comparison of position 2 at 6.7m, Fig. 5 over position 1 at 6.7m, Fig. 5C over position 2 at 3.35m, and Fig. 5D over position 1 at 3.35m. The dashed and solid lines show the measurements in the 90 and 45 degree orientations respectively. Fig. 5A shows good agreement between the two orientations. The other three charts have some channels showing good agreement and some channels showing large differences. These differences can most likely be attributed to wind noise contaminating the measurements. D
A Normalized Pressure (unitless) measurement of the 3.35m position was louder than the corrected SPL measurement in the 6.7m position. If the ratio has a negative value, the SPL measurement in the 3.35m position was quieter than the corrected SPL measurement in the 6.7m position. In general, the ratios are close to zero, however channel 9 from Fig. 6A and channel 2 from Fig. 6 have values significantly larger than 0. This was most likely due to Excessive wind noise in the measurements. 4. Conclusion Normalized Pressure (unitless) Fig. 6A and 6 show a normalized and averaged comparisons between the higher positions and the lower positions in position 1 and 2 respectively. The pressure signal was A weighted to reduce the effect of wind noise in the measurement The normalized, corrected ratios of the lower positions to the higher position is calculated using the following: Eq.3 R = P! P! C 1 Where P l is the averaged root-mean square pressure (Prms) of each microphone channel in the lower position of 3.35m, P h is the averaged Prms of each microphone channel in the higher position of 6.7m, C is the appropriate geometric correction factor for each microphone channel. If the UAV were to act as an omnidirectional point source, the ratio of!! will equal 1, so subtracting 1 from!!! the ratio will center the chart around zero. If the ratio has a positive value, the SPL Accurate and controlled UAV noise measurements are necessary for aiding the integration of UAVs into everyday urban life. Accessing the characteristics of the noise and understanding its conception is imperative for reducing noise levels and improving UAV design and efficiency. For the UAV tested, strong harmonic tendencies are observed, stemming from the blade-pass frequency of each propeller. Above 5000Hz, the signal becomes noise, containing all frequencies with the amplitude decreasing with increasing frequency. Every odd numbered harmonic of the fundamental frequency of the signal is amplified, most likely caused by the orientations of the propellers. Due to the physical constraints of the anechoic chamber, the maximum height the UAV could safely hover above the microphone array was 6.71m. This distance was not sufficient to minimize wind turbulence interference. For future measurements, it is suggested that the UAV be flown more than 10m away from a recording source to avoid wind turbulence interference. Acknowledgements This project was made in collaboration with Dr. Timothy rungart, Dr. Steven Olson, and Zachary Yoas of the Applied Research Laboratory at The Pennsylvania State University. Additionally, special thanks is given to Dr. Stephen Thompson for advising the author throughout the course of this project.
References and Links Hubbard, Harvey. Aeroacoustics of Flight Vehicles: Theory and Practice. Nasa, Washington, United States: Vol. 1. (Aug 1991). Marte, Jack. Kurtz, Donald. A review of Aerodynamic Noise From Propellers, Rotors, and Lift Fans. Nasa Technical Report. Jet Propulsion Laboratory, California Institute of Technology, Pasadena, California. (Jan 1970). Staples, Gabriel. Propeller Static & Dynamic Thrust Calculation. < http://www.electricrcaircraftguy.com/2014/04/propeller-static-dynamic-thrustequation-background.html >. Accessed (July 2017). Google Trends unmanned aerial vehicle search, < https://trends.google.com/trends/explore?date=all&q=%2fm%2f0g2bc>. Accessed (July 2017).