Optimization of a contra-rotating propeller rig for reduced psychoa- coustic impact Fabio Casagrande Hirono 1 , James Robertson 2 , Antonio Torija Martinez 3 , Andrew Elliott 4 , Acoustics Research Centre, University of Salford 43 Crescent, Salford M5 4WT

ABSTRACT With unmanned aerial vehicles emerging as potential alternatives for people and cargo transport, their noise impact will be a determining factor in their acceptance by the general public. Contra- rotating propeller configurations are often explored due to their improved aerodynamic performance and redundancy in case of failure compared to conventional single-propeller aircraft, but can be much noisier than their single-propeller equivalent. This work describes the optimization of a custom- made contra-rotating propeller rig for reduced psychoacoustic impact. The rig consists of two elec- tric motors mounted on a rotating stand, positioned inside an anechoic chamber. A far-field micro- phone arc is used to collect acoustic pressure data, and a load cell is used to measure total thrust. The axial distance between the propellers is varied between 0.1 and 1 rotor diameters, and the num- ber of blades is varied between 2 and 6 on both propellers. Meanwhile, the rpm is adjusted to main- tain constant thrust across the different configurations. Acoustic pressure signals are investigated in terms of their physical acoustic characteristics and psychoacoustic features (such as Loudness, To- nality, and Impulsiveness) in order to determine the trade-offs and optimal choices in reducing the psychoacoustic impact of the rig.

1. INTRODUCTION Unmanned aerial vehicles (UAVs) have become a promising future alternative to urban transport, and it is generally agreed that public acceptance of such novel aircraft will be strongly influenced by their noise impact. Multi-rotor architectures are particularly appealing for their compact size and im- proved performance compared to a single-rotor aircraft, but can be notoriously noisy when not care- fully designed (McKay et al, 2019).

Contra-rotating propeller configurations have been previously investigated in the literature for their improved aerodynamic performance, reduced planform area and added redundancy in case of component failure. Contra-rotating propellers generate a mixture of tonal and broadband noise, with the tonal components dominating the spectrum at low to mid frequencies and decreasing in amplitude at higher frequencies, and the broadband component dominating the spectrum at higher frequencies

1 F.CasagrandeHirono@salford.ac.uk

2 J.Robertson6@edu.salford.ac.uk

3 A.J.TorijaMartinez@salford.ac.uk

4 A.S.Elliott@salford.ac.uk

worm 2022

(Parry et al, 2011). Each individual rotor generates tonal noise, composed of thickness and (steady) loading noise, and broadband noise, consisting of turbulence-interaction (i.e. leading edge) and trail- ing edge noise. In addition, significant tonal noise is generated at multiple sum and difference tones by potential field interaction between both rotors when these are closely spaced (Parry et al, 2011).

Potential field interaction tones are stronger on-axis and can be around 20 dB higher than rotor-alone tones at typical observer locations (McKay et al, 2019). These tones clearly dominate the spectrum at short axial distances but decrease exponentially in magnitude with increasing axial spac- ing (McKay et al, 2019; Chaitanya et al, 2021). On the other hand, tonal noise sources due to indi- vidual rotors are not expected to change significantly with varying axial distance.

Turbulence shed by the upstream rotor will interact with the downstream rotor and generate broadband noise. Blandeau and Joseph (2010) observed a significant reduction in rotor-wake/rotor interaction noise with decreasing axial distance due to decreasing wake spreading. In a later study, Blandeau et al (2013) showed that trailing edge noise from each rotor is unaffected by the rotor axial spacing.

Chaitanya et al (2021) investigated the noise generation mechanisms of two-bladed overlapping propellers at different axial spacings, and proposed an axial separation distance of 0.25 rotor diame- ters as an optimum balance between the contribution of the various acoustic sources, leading to min- imum noise impact. Torija et al (2021) used Sound Quality Metrics (SQMs) to investigate the optimal rotor axial distance for two-bladed contra-rotating propeller for minimum psychoacoustic impact, and identified the range of axial distances between 0.2 and 0.4 rotor diameters as optimal for reduced psychoacoustic impact.

This paper describes a series of experiments investigating the optimization of a contra-rotating propeller rig for reduced psychoacoustic impact. We extend the analyses previously made by Torija et al (2021) to include the effect of varying number of blades as an extra design parameter. Both rotors were assembled with identical number of blades and operated at the same nominal thrust value at multiple rotor axial distances. We report acoustic and psychoacoustic results for an observer at 180° (i.e. on-axis, away from the mean flow), where potential field interaction tones should dominate the spectrum.

2. EXPERIMENTAL SETUP

The acoustic measurements of the contra-rotating rig were performed in the University of Salford anechoic chamber, as shown in Figure 1. The anechoic chamber has a working space of 5.4 x 4.1 x 3.3 m, with a background noise level of -12.4 dB(A) and a cut-off frequency of 100 Hz (University of Salford, 2022). 2.1. Contra-rotating test rig The contra-rotating test rig consisted of two Turnigy SK3 Aerodrive 3542 800 motors mounted on a steel and aluminium frame, and is shown in Figure 1. A set of custom-made hubs were used to attach between two to six blades to each motor. The blades were off-the-shelf FlightLine 12x7 , in both forward and reverse configuration. The blade tips were trimmed for use in a separate experiment, resulting in a rotor diameter D = 0.28 m. The motors were attached to axles, which in turn were attached to moving stands mounted on railings. The moving stands allowed for changes in the rotor-rotor axial distance z , which could be adjusted between 0.1 D and 1.0 D . Measurements were performed for axial spacings of 0.1, 0.2, 0.3, 0.4, 0.6, 0.8, and 1.0 rotor diameters. The contra-rotating rig was mounted on a rotating stand at the centre of

worm 2022

the anechoic chamber, thus allowing for measurements at different azimuthal angles, and the stand was instrumented with a 10 kg load cell for measuring the total thrust of the rig. The load cell was calibrated by applying a known force in the direction of positive thrust and calibrating the resulting signal in the data acquisition system. As changes to the moving stand positions led to changes in the mass distribution of the rig, the load cell signal was rebalanced to compensate for its new static load- ing after any changes to the rotors’ axial distance.

worm 2022

Load Cell

Figure 1: Contra-rotating rig positioned on a rotating stand in the University of Salford anechoic chamber. 2.2 Data acquisition system The acoustic data acquisition system consisted of a fixed microphone arc with ten free-field, ½” mi- crophones at R= 2.5 m radius. The microphones were installed at polar angles between θ= 0° (i.e. on- axis, downstream) to 90° (on the propeller plane), in 10° steps. Further measurements of polar angles between 90° and 180° (on-axis, upstream) could be performed by rotating the rig stand by 180° in the azimuthal direction. The microphones used in the arc were seven B&K 4966-H-041 free-field micro- phone sets, and three 01dB MCE212 free-field microphone capsules connected to G.R.A.S. 26CA preamplifiers. All microphones were calibrated using a 1 kHz, 94 dB SPL B&K calibrator. As some measurements were performed with the rig exhaust directed towards the microphones, the three low- est microphones were equipped with wind shields.

Figure 2: Contra-rotating rig and microphone arc inside anechoic chamber. The polar angle θ =0° denotes the downstream direction, and θ =180° denotes the upstream direction. The microphones and the load cell were connected to a rack of Dewesoft Sirius data acquisition units, and their signals were recorded to a laptop computer using the proprietary DewesoftX data acquisition software. The microphone signals were sampled at f s = 50 kHz, whereas the load cell signal was sam- pled synchronously to the acoustic signals but at f s2 = 12.5 kHz instead. A real-time monitoring screen was set up in DewesoftX to observe the thrust and acoustic signals in real time. The motors’ rotational speed were manually controlled using a single HIJ Digital servo controller connected to two Maytech MT40A-OPTO-SF32 Electronic Speed Controllers (ESCs), powered from a Turnigy Reaktor Pro 350W power supply. The rotational speeds varied slightly between the two rotors, likely due to small imperfections in the motors and rotor blades used. Both rotors were assembled with an equal number of blades (between 2 and 6) and positioned at the desired axial distance z . Their rotational speed was adjusted using the servo controller until the desired thrust value was observed in the real-time monitoring screen. Measurements were performed for a duration of 30 s with the microphones downstream of the rig to obtain the noise at θ =0° to 90°. The rig was then rotated 180° in the azimuthal direction, and new measurements were performed with the microphones upstream of the rig to obtain the noise at θ =90° to 180°. This procedure was repeated for two thrust values (4 N and 8 N), for all combinations of blade numbers and axial spacing. The microphone signals were processed in Matlab using Welch’s method to obtain their narrowband Power Spectral Densities (PSDs) using a DFT size of 2 16 samples and a Hann window function, yielding a frequency resolution of 0.7629 Hz. The broadband component of the PSD was estimated by applying a moving-median filter to the spectra (Parry et al, 2011), and the tonal components of the PSD were estimated by manually annotating the peak locations in the PSDs. Sound Pressure Levels (SPLs) were obtained by integrating each PSD components within the frequency range [100, 5000] Hz, resulting in separate broadband SPL, tonal SPL, and total (broadband plus tonal) SPL for each configuration.

worm 2022

In addition to the acoustic analysis above, the raw signals acquired from the microphones were ex- ported as .WAV files and processed in HEAD Acoustics ArtemiS 12.0 to obtain a series of Sound Quality Metrics (SQMs) as a function of time. In this work, we focus on Loudness (DIN 45631/A1), Tonality (Aures model), and Impulsiveness (Hearing model). The 5 th percentile of each SQM time series was calculated and is used to represent the higher range of a SQM time history. To avoid initial transients in the SQM calculations, the initial 0.5 seconds of each time history was discarded.

3. RESULTS

In this paper we focus on acoustic measurements taken at θ =180° (i.e. on-axis, upstream) only, and with 2, 4, or 6 blades on each rotor. Figures 3, 4, and 5 show narrowband frequency spectra between 20Hz to 5kHz for all blade numbers, both thrust settings, and select axial spacings. This frequency range is chosen as this is where the majority of tonal peaks occur.

As both rotors were driven at nominally identical rotational speed and equipped with the same number of blades, their BPF harmonics and interaction tones significantly overlap across the fre- quency spectra. This is indicated in Figure 3(a), where the tones’ frequencies are annotated with their corresponding harmonic indices { n 1 , n 2 } for {BPF1, BPF2}. For example, the second harmonics {2,0} and {0,2} overlap with the first interaction tone {1,1}. The tone amplitudes decrease with in- creasing frequency at approximately the same rate in all cases, as noted by McKay et al (2019), and very little tonal noise is present in the spectra for frequencies above 5 kHz.

It can be observed that potential field interaction tones where n 1 = n 2 have significantly higher magnitudes than other BPF and interaction tones at this direction, as established in the literature for rotors with identical number of blades (McKay et al, 2019). The broadband noise component is seen to increases with increasing axial spacing in all configurations, pointing towards a likely predomi- nance of turbulence-interaction noise versus trailing edge noise in this frequency range. 3.1 Noise levels vs. axial spacing Figures 6 and 7 show the variations in tonal, broadband, and total (tonal plus broadband) SPL as a function of axial spacing for various number of blades and thrust settings. The total noise shows highest magnitudes at short axial spacing, decreases significantly for z/D values in the 0.3-0.4 range, and generally increases by a few decibels for larger spacing. Hence, an optimal distance can be found for minimum noise emissions.

Tonal noise due to potential field interaction tones dominate the emissions at very short axial spacing at all conditions. Short axial spacings also correspond to the lowest levels of broadband noise in all conditions, and therefore the largest ratio of tonal-to-broadband noise levels is obtained at small z/D . As axial distance increases, tonal noise decreases abruptly, reaching approximately steady levels for axial spacing z/D ≥ 0.6.

On the other hand, broadband noise steadily increases with increasing axial distance, and eventually surpasses tonal noise levels around z/D = 0.3 to become the dominant noise source. There- fore, an optimum balance between the two types of noise can be found for axial spacing near z/D = 0.3, as previously described in the literature (Chaitanya et al, 2021; Torija et al, 2021).

worm 2022

worm 2022

Figure 3 (a) (left) & (b) (right): 4 N (left) and 8 N (right) PSD spectra plots for 2 blades and different axial spacing. The indices { n 1 , n 2 } indicate the BPF harmonic indices for BPF1 and BPF2 .

42.2), (1.3), (3.1), (0.4), (40) 40.2) 2.0) GA), (5.1), 1.5), fy (42), (2.4) (06), (6.0)

Figure 4 (a) (left) & (b) (right): 4 N (left) and 8 N (right) PSD spectra plots for 4 blades and different axial spacing.

Frequency Ha) Frequency (Ha)

Figure 5 (a) (left) & (b) (right): 4 N (left) and 8 N (right) PSD spectra plots for 6 blades and different axial spacing.

we we we ow "| Frequency (H2)

3.2 Noise levels vs. number of blades By keeping thrust constant, increasing the number of blades is expected to reduce the steady loading on each blade, as the same force is distributed over more blades. Consequently, the same thrust can be achieved with an increased blade count at a smaller rotational speed, although at the cost of in- creased drag and reduced aerodynamic efficiency. Increasing the blade count also increases the sep- aration between subsequent BPF harmonics, and consequently between potential field interaction tones, as seen in Figures 3 to 5. Figures 6 and 7 show that increasing the number of blades at constant thrust results in a small decrease of tonal noise at larger axial spacing, possibly due to the reduced static loading and blade tip speeds involved at higher blade counts. This is accompanied by a much larger increase in potential field interaction tonal noise at short distances, as seen by the increased interaction tones’ magnitudes across Figures 3 to 5 for z/D =0.1. Increasing the number of blades is also shown to slightly decrease the broadband noise levels across all spacings, likely due to the reduced blade tip speed as well.

Considering the points discussed above, it is seen that increasing the number of blades while keeping thrust equal in a contra-rotating propeller system significantly increases tonal noise at small axial spacing, while slightly decreasing both broadband and tonal noise at higher axial spacings. 4. Psychoacoustic analysis We now discuss how the above findings compare to psychoacoustic analysis of the same recordings. Figure 8 shows the computed Sound Quality Metrics as a function of axial spacing for both thrust settings and all blade numbers. The loudness plots show very similar trends to the Total SPL plots in Figures 6 and 7, with high values for low axial spacing and low values otherwise. As loudness describes the perception of sound intensity, this is expected (Head Acoustics, 2018). A point of minimum loudness is visible around z/D =0.3 in almost all conditions, whereas in some cases the loudness remains at a low value for larger axial spacing. It can also be observed that loudness significantly increases in magnitude for increasing thrust, as all noise sources become louder at higher rotational speeds. However, only small variations in loudness are observed for different number of blades, with less blades being slightly louder. This is in line with the observations made previously from the acoustic metrics. The plots for tonality follow a similar trend to loudness, with high values at low axial spacing and decreasing at higher spacing (Torija et al, 2021). As previously observed, the ratio between tonal and broadband noise is also significantly higher at shorter axial spacing and decreases with increasing z/D , thus explaining the behaviour shown below. No significant changes in tonality are observed with increasing thrust, and only a small increase can be noted for decreasing number of blades. It is to be noted that the calculation of Aures tonality is complex and accounts for the position of the tones, the number of tones, and prominence over masking sound level within the critical band.

worm 2022

worm 2022

Figure 6: Tonal, broadband, and total SPL as a function of axial spacing for all blade numbers, at 4N thrust.

2 blades, 8N blades, 8N 4 blades, 8N 0 Py 0 ts ts ts | * o®| 4 a” 8 Bi 8 S75 pecctcct Saf & * Sis t z wepececsict gt ' : 27) \ eno 5] See | ee | acho os os ts oo « 002 os a6 os 1 % 2 02 06 of 1 > oz os o6 o8 1+ Axial spacing, 2/D. Axial spacing, z/D Axial spacing, 2/D. °0, 5 blades, 8N © 6 blades, 8N ‘ se a) a) ao) “+ Tonal noise 100-5kHz 8 S » Broadband noise 100-SkHz 2” age gy RY Total noise (BB+Tonal) 100-SkHz Dx PSS Bp oe ee os| 5 . . + fo a2 oa 06 08 1 oo aa 08 08 + Axial spacing, 2/D Axial spacing, 2/D

Figure 7: Tonal, broadband, and total SPL as a function of axial spacing for all blade numbers, at 8N thrust. Note the different vertical scale compared to Figure 6.

2 blades, 4N ‘blades, 4N 4 blades, 4N 6 wy 4 * \ Br Br weet a iS | 4 5 Beeecte oS Bes 7 Be et wt al ott © en 5 55 02 04 06 08 ‘Axial spacing, 2/D 002 04 06 08 Axial spacing, 2/D 5 blades, 4N ~ 6 blades, 4N 1 ‘ wl \ \ Bn) iin t ae ee ae et al ae cd ss 02 0s 08 08 Axial spacing, 2/D 002 04 08 08 Axial spacing, 2/D 002 04 06 08 31 Axial spacing, 2/D “+ Tonal noise 100-SkHz Broadband noise 100-SkHz * Total noise (BB+Tonal) 100-5kHz

Figure 8: Loudness, Tonality, and Impulsiveness Sound Quality Metrics for 2, 4, and 6 blades versus axial spacing, at 4N thrust (left) and 8N (right).

worm 2022

The impulsiveness plots show very low values at small axial distances, and increase with increas- ing axial spacing for most conditions. Large differences in impulsiveness are observed for different blade numbers, with 6-bladed configurations showing reduced impulsiveness for both thrust settings and almost all axial spacings. At higher thrust settings, the blade count has less effect on impulsive- ness. Currently these trends are not completely understood and require further research.

Given the opposing trends between both loudness and tonality versus impulsiveness as functions of axial spacing, it can be proposed that an optimum axial spacing exists for minimum psychoacoustic impact by balancing the contribution of the individual Sound Quality Metrics. This minimum is likely located near z/D =0.3, which coincides with the acoustic analysis reported above and in the literature (Chaitanya et al, 2021; Torija et al, 2021).

4. CONCLUSIONS

This work investigated the acoustic and psychoacoustic characteristics of a contra-rotating rig for an observer at θ =180°, and for a range of blade numbers, rotor axial spacing, and thrust settings. Good agreement was obtained with results previously reported in the literature.

The main conclusions are: • Increasing the number of blades while keeping thrust equal in a contra-rotating propeller sys-

tem significantly increases tonal content at small axial spacing, but slightly decreases both broadband and tonal components at higher axial spacings; • Loudness and tonality do not change significantly with varying blade number, whereas im-

pulsiveness shows lower values at higher blade numbers; • The blade count has less impact on the impulsiveness at higher rotational speeds (i.e. higher

thrust setting);

2-2 blades ~* 4 blades 6-6 blades Loudness DIN45631, 4N 02 04 06 os 1 Tonality AURES, 4N 02 04 06 08 1 s impulsiveness (Hearing Mode), 4N Zon g Boa 3 Zo = fo 02 04 06 08 1 ‘Axial Spacing, 2/D Loudness DIN45631, 8N 02 04 06 os 1 Tonality AURES, 8N ° 02 04 06 08 1 02 04 06 08 1 ‘Axial Spacing, 2/D

• Based on both acoustic and psychoacoustic results, the optimal spacing for a contra-rotating

rotor system for minimum noise impact is around 0.3 z/D for all blade counts, corroborating previous results in the literature. For future work, the psychoacoustic trends presented in this report will be validated through a participant listening experiment. A more complete psychoacoustic annoyance study combining mul- tiple SQMs could be beneficial to better understand annoyance trends of contra-rotating rotor systems design. Such metrics have been previously proposed by Torija et al (2021; 2022) and will be investi- gated in future work. 5. ACKNOWLEDGEMENTS

The Authors would like to acknowledge the funding provided by the UK Engineering and Physical Sciences Research Council for the DroneNoise project (EP/V031848/1), and by Innovate UK for the InCEPTion project (ref. 73692). 6. REFERENCES

Blandeau, V., Joseph, P. (2010) ‘Broadband Noise Due to Rotor-Wake/Rotor Interaction in Contra- Rotating Open Rotors’, in AIAA Journal 48 (11). Available at: https://arc.aiaa.org/doi/10.2514/1.J050566

Blandeau, V., Joseph, P., Kingan, M., Parry, A. (2013) ‘Broadband Noise Predictions from Unin- stalled Contra-Rotating Open Rotors’, in Int. J. Aeroacoustics 12 (3). Available at: http://jour- nals.sagepub.com/doi/10.1260/1475-472X.12.3.245

Brentner, K., Farassat, F. (2003) ‘Modeling aerodynamically generated sound of helicopter rotors’, in Progress in Aerospace Sciences 39. Available at: https://doi.org/10.1016/S0376-0421(02)00068-4

Chaitanya, P., Joseph P., Akiwate, D., Parry, A.B., and Prior, S.D. (2021) ‘On the noise generation mechanisms of overlapping propellers’, in. AIAA Aviation 2021 Forum , Virtual. Available at: https://arc.aiaa.org/doi/10.2514/6.2021-2281 HEAD Acoustics (2018) 'Loudness and sharpness calculation', Technical Report Application Note 02/18, HEAD Acoustics, Brighton, MI. Available at: https://cdn.head-acoustics.com/filead- min/data/global/Application-Notes/SVP/Psychoacoustic-Analyses-I-02.2018.pdf (Accessed: 15 Feb- ruary 2022). McKay, R.S., Kingan, M.J. and Go, R. (2019) ‘Experimental investigation of contra-rotating multi- rotor UAV propeller noise’, in. Proceedings of Acoustics 2019 , Cape Schanc, Victoria, Australia. Available at: https://acoustics.asn.au/conference_proceedings/AAS2019/papers/p20.pdf (Accessed: 4 April 2022).

Parry, A., Kingan, M., Tester, B. (2011) 'Relative importance of open rotor tone and broadband noise sources', in 17 th AIAA/CEAS Aeroacoustics Conference (32 nd AIAA Aeroacoustics Conference) , Port- land, Oregon, USA. Available at: https://arc.aiaa.org/doi/10.2514/6.2011-2763

worm 2022

Torija, A.J., Chaitanya, P. and Li, Z. (2021) ‘Psychoacoustic analysis of contra-rotating propeller noise for unmanned aerial vehicles’, The Journal of the Acoustical Society of America , (149), pp. 835–846.

Torija, A. J., Li, Z., Chaitanya, P. (2022) 'Psychoacoustic modelling of rotor noise', Journal of the Acoustical Society of America (151). Available at: https://doi.org/10.1121/10.0009801 University of Salford (2022), ‘Anechoic Chamber’. Available at: https://acoustictesting.sal- ford.ac.uk/acoustic-laboratories/anechoic-chamber/

worm 2022