What makes drones sound annoying? The answer may lie in noise fluctuations

Ze Feng (Ted) Gan – tedgan@psu.edu

Department of Aerospace Engineering, The Pennsylvania State University, University Park, PA, 16802, United States

Popular version of 2aNSa3 – Multirotor broadband noise modulation
Presented at the 186th ASA Meeting
Read the abstract at https://eppro02.ativ.me/web/index.php?page=IntHtml&project=ASASPRING24&id=3673871

–The research described in this Acoustics Lay Language Paper may not have yet been peer reviewed–

Picture yourself strolling through a quiet park. Suddenly, you are interrupted by the “buzz” of a multirotor drone. You ask yourself: why does this sound so annoying? Research demonstrates that a significant source is the time variation of broadband noise levels over a rotor revolution. These noise fluctuations have been found to be important for how we perceive sound. This research has found that these sound variations are significantly affected by the blade angle offsets (azimuthal phasing) between different rotors. This demonstrates the potential for synchronizing the rotors to reduce noise: a concept that has been studied extensively for tonal noise, but not broadband noise.

Sound consists of air pressure fluctuations. One major source of sound generated by rotors consists of the random air pressure fluctuations of turbulence, which encompass a wide range of frequencies. Accordingly, this sound is called broadband noise. A common example and model of broadband noise is white noise, shown in Figure 1, where the random nature characteristic of broadband noise is evident. Despite this randomness, we hear the noise of Figure 1 as having a nearly constant sound level.

Figure 1: White noise with a nearly constant sound level.

A better model of rotor noise is white noise with amplitude modulation (AM). Amplitude modulation is caused by the blades’ rotation: sound levels are louder when the blade moves towards the listener, and quieter when the blade moves away. This is called Doppler amplification, and is analogous to the Doppler effect that shifts sound frequency when a sound source travels towards or away from you. AM white noise is shown in Figure 2: the sound is still random, but has a sinusoidal “envelope” with a modulation frequency corresponding to the blade passage frequency. AM causes time-varying sound levels, as shown in Figure 3. This time variation is characterized by the modulation depth, the peak-to-trough amplitude in decibels (dB), as shown in Figure 3. A greater value for modulation depth typically corresponds to the noise sounding more annoying.

Figure 2: White noise with amplitude modulation (AM).
Figure 3: Time-varying sound levels of AM white noise.

Broadband noise modulation is known to be important for wind turbines, whose “swishing” is found to be annoying even at low sound levels. This contrasts with white noise, which is typically considered soothing when it has a constant sound level (i.e., no AM). This exemplifies the importance of considering time variation of sound levels for capturing human perception of sound. More recently, the importance of broadband noise modulation has been demonstrated for helicopters, as this chopping noise is what makes a helicopter sound like a realistic helicopter, even if it has low sound levels.

Researchers have not extensively studied broadband noise modulation for aircraft with many rotors. Computational studies in the literature predict that summing the broadband noise modulation of many rotors causes “destructive interference”, resulting in nearly no modulation. However, flight test measurements of a six-rotor drone showed that broadband noise modulation was significant. To investigate this discrepancy, changes in modulation depth were studied as the blade angle offset between rotors was varied. This offset is typically not considered in noise predictions and experiments. The results are shown in Figure 4. For each data point in Figure 4, the rotor rotation speeds are synchronized, but the value for the constant blade angle offset between rotors is different. The results of Figure 4 demonstrate the potential for synchronizing rotors to reduce broadband noise modulation. This synchronization controls the blade angle offset between rotors to be as constant as possible, and has been extensively studied for controlling tones (sounds at a single frequency), but not broadband noise modulation.

Figure 4: Modulation depth as a function of blade angle offset between two synchronized rotors.

If the rotors are not synchronized, which is typically the case, the flight controller continuously varies the rotors’ rotation speeds to stabilize or maneuver the drone. This causes the blade angle offsets between rotors to with vary with time, which in turn causes the summed noise to vary between different points in Figure 4. Measurements showed that all rotor blade angle offsets are equally likely (i.e., azimuthal phasing follows a uniform probability distribution). Therefore, multirotor broadband noise modulation can be characterized and predicted by constructing a plot like Figure 4, by adding copies of the broadband noise modulation of a single rotor.

The loss of an F35 fighter aircraft and the search for Malaysian Airlines flight MH370

Alec Duncan – a.j.duncan@curtin.edu.au

Centre for Marine Science and Technology, Curtin University, Bentley, WA, 6102, Australia

David Dall’Osto
Applied Physics Laboratory
University of Washington
Seattle, Washington
United States

Popular version of 1pAO2 – Long-range underwater acoustic detection of aircraft surface impacts – the influence of acoustic propagation conditions and impact parameters
Presented at the 185th ASA Meeting
Read the abstract at https://doi.org/10.1121/10.0022761

Please keep in mind that the research described in this Lay Language Paper may not have yet been peer reviewed.

In the right circumstances, sound can travel thousands of kilometres through water, so when Malaysian Airlines flight MH370 went missing in the Indian Ocean in 2014 we searched recordings from underwater microphones called hydrophones for any signal that could be connected to that tragic event. One signal of interest was found, but when we looked at it more carefully it seemed unlikely to be related to the loss of the aircraft.

Fast-forward five years and in 2019 the fatal crash of an F35 fighter aircraft in the Sea of Japan was detected by the Comprehensive Nuclear-Test-Ban Treaty Organisation (CTBTO) using hydrophones near Wake Island, in the north-western Pacific, some 3000 km from the crash site1.

Fig. 1. Locations of the F35 crash and the CTBTO HA11N hydroacoustic station near Wake Island that detected it.

With the whereabouts of MH370 still unknown, we decided to compare the circumstances of the F35 crash with those of the loss of MH370 to see whether we should change our original conclusions about the signal of interest.

Fig. 2. Location of the CTBTO HA01 hydroacoustic station off the southwest corner of Australia. The two light blue lines are the measured bearing of the signal of interest with an uncertainty of +/- 0.75 degrees.

We found that long range hydrophone detection of the crash of MH370 is much less likely than that of the F35, so our conclusions still stand, however there is some fascinating science behind the differences.

Fig. 3. Top: comparison of modelled received signal strengths versus distance from the hydrophones for the MH370 and F35 cases. Bottom: water depth and deep sound channel (DSC) axis depth along each path.

Aircraft impacts generate lots of underwater sound, but most of this travels steeply downward then bounces up and down between the seafloor and sea surface, losing energy each time, and dying out before it has a chance to get very far sideways. For long range detection to be possible the sound must be trapped in the deep sound channel (DSC), a depth region where the water properties stop the sound hitting the seabed or sea surface. There are two ways to get the sound from a surface impact into the DSC. The first is by reflections from a downward sloping seabed, and the second is if the impact occurs somewhere the deep sound channel comes close to the sea surface. Both these mechanisms occurred for the F35 case, leading to very favourable conditions for coupling the sound into the deep sound channel.

Fig. 4. Sound speed and water depth along the track from CTBTO’s HA11N hydroacoustic station (magenta circle) to the estimated F35 crash location (magenta triangle). The broken white line is the deep sound channel axis.

We don’t know where MH370 crashed, but the signal of interest came from somewhere along a bearing that extended northwest into the Indian Ocean from the southwest corner of Australia, which rules out the second mechanism, and there are very few locations along this bearing where the first mechanism would come into play.

Fig. 5. Sound speed and water depth in the direction of interest from CTBTO’s HA01 hydroacoustic station off Cape Leeuwin, Western Australia (magenta circle). The broken white line is the deep sound channel axis.

This analysis doesn’t completely rule out the signal of interest being related to MH370, but it still seems less likely than it being due to low-level seismic activity, something that results in signals at HA01 from similar directions about once per day.

[1] Metz D, Obana K, Fukao Y, “Remote Hydroacoustic Detection of an Airplane Crash”, Pure and Applied Geophysics,  180 (2023), 1343-1351. https://doi.org/10.1007/s00024-022-03117-6