Self active cancellation of fan noise

Abstract

Reduction of fan noise is an important problem in the successful deployment of drones and UAV's. This paper considers a new approach to reducing fan and propeller noise based upon micro vibrations of the propeller blades around their axis of support. Experimental testing was carried out on a five bladed fan arrangement. The micro fan blade vibrations are induced with a pitch link actuator arrangement driven by an electromagnetic actuator. When used in conjunction with a digital feedforward active noise controller, the micro blade vibrations were found to provide global attenuations of fan radiated sound the order of 5 to 10dB of the first few fan tones. The level of required vibrations and the associated electrical power required for the cancelling micro vibrations was shown to be very small compared to the fan motor power requirements. The results demonstrate that the innovative approach, termed “self active cancellation of fan noise”, has good potential for global reduction of fan and propeller noise.

Keywords

Fan noise active noise control smart blades

Introduction

In the near future it very likely there will be high numbers of drones and UAV’s in commercial, military and private use in the US. While drones bring many benefits, one of the concerns is the noise produced by the multiple propeller systems employed for lift and control and this problem will very likely reduce their successful deployment, particularly in crowded urban areas. It is very likely that drone noise will be the biggest barrier to their general acceptance in society. This acoustic problem is exacerbated by the fact that the propellers are contained in very short ducts(or no ducts) and this effectively stops the use of acoustic absorptive materials to reduce the radiated noise as in ducted turbo fan engines. Shaping of the blades to reduce noise has already been pursued to the maximum and has the down side of also effecting fan aerodynamic performance. As such alternative novel methods need to be found to control the noise. Active noise cancellation(ANC) has been suggested as a potential solution. The usual method of implementing ANC is to employ an array of active acoustic speakers to generate a sound cancelling field and this has been investigated and demonstrated by Prof. Fuller and his team in the active control of noise radiated out of the inlet of turbo fan engines¹ and for ducted computer fans.² As shown by Prof. Fuller the application of these types of ANC system require a high number of active speakers and an extended duct either side of the fan in which to install the active speakers. This conventional ANC approach is thus very likely to be too heavy and bulky for the drone fan application and is thus not suitable.

Concept of self active noise reduction

In this approach to the active control of drone fan noise, it is proposed to vibrate the fan blades around their own axis to create an active cancellation sound field that will “self cancel” the actual fan tones. To support this concept we have carried out some preliminary experiments as described below. It has been demonstrated by a number of researchers and manufacturers that actuating the fan blades around their individual axes while the fan is rotating will produce sound that is coherent with the oscillating signal used to actuate the blades. The sound is produced by the time varying modulation of the air flow over the blades much like a siren. These modulations are very small in amplitude but the fan source can produce very large levels of sound at low frequencies. For example the TRW 17 fan noise source shown in Figure 1 produced by Eminent Technologies ³ using a mechanical linkage to actuate the fan blades around their axes while the fan is rotating at a set speed. Its output sound spectrum normal to the fan at 1m for a 1 Watt drive signal is also shown in Figure 1. The general concept of the self ANC system proposed here is to obtain a reference signal from the fan shaft motion which will be coherent with the tones radiated by the fan. This signal will be then passed through a digital feedforward controller and used to derive a control signal to drive the vibration of the fan blades in order to globally (throughout an extended space) cancel the fan tones. for a white noise drive signal.

Figure 1.

Eminent Technologies TRW 17 infra sound source and its output spectrum for a white noise drive signal.³

Prof. Fuller and the Center for Aerospace Acoustics have carried out experiments to demonstrate that these types of fan source can be used to actively cancel sound and this is discussed below. A digital adaptive feedfoward control system implementing the Filtered X algorithm was used as the controller.⁴ The controller used the noise drive signal as an ideal reference signal to generate a signal coherent with the fan noise and then produced a control signal which was used to drive the voice coil which actuated the fan blades. The controller then adaptively adjusted the control signal to minimize the sound at the error microphone in the far field. Figure 2 shows a block diagram of a typical SISO feedfoward control arrangement.

Figure 2.

Block diagram of a SISO feedfoward control arrangement.

As discussed in the review paper on ANC by Fuller and Von Flotow,⁵ feedfoward control implementations are eminently suitable for applications where a reference signal, coherent with the noise, is readily available. For fan noise this is easily obtainable from a shaft rotational sensor whose signal has been shown in many applications to be coherent with the tones produced by propellers and fans.^1,2 In Figure 2, x(n) is the disturbance signal at time step n, P(z) is the plant in the z-domain, d(n) is the plant response(uncontrolled) at time step n, W(n) is the adaptive filter, S(z) is transfer function from the control actuator input to the error sensor output, e(n) is the error sensor signal at time step n, y(n) is the controller output signal at time step n and x'(n) is the filtered reference signal used in the LMS^4,5 update equation used to update the adaptive filter coefficients at each time step n. The LMS update equation for a SISO arrangement is,^4,5

w (n + 1) = w (n) - μ e (n) x' (n)

(1)

where µ is the stability gain factor and w(n+1) are the new adaptive filter coefficients at the next time step. The last term in equation (1) is the instantaneous gradient of the quadratic cost function formed from e(n).² Thus the LMS algorithm calculates the instantaneous gradient of the cost function at time n. It then uses this gradient to calculate the new adaptive filter coefficients by stepping down the cost function in the negative direction of the gradient towards the cost function minimum. Note, in the block diagram everything except P(z) the physical plant is usually implemented in digital fashion on an appropriate DSP system. See the text by Fuller et al.³ for more details on this algorithm and its extension to MIMO.

Previous related work

In related work, Noctua, a Swedish company, has produce an ANC for controlling computer fan noise using a similar approach.⁶ In the Noctua system, the fan blades are actuated around their axes by small magnets embedded in the blade tips and an electrical coil embedded in the fan housing as shown in Figure 3(a). Using this arrangement in conjunction with a controller, Noctua has demonstrated ANC of the fan noise tones as shown in Figure 3(b). In other words the Noctua system uses small oscillations of the fan blades around their individual axes to self cancel the fan noise itself. This is a very exciting development and has direct application to successful ANC of UAV noise.

Figure 3.

(a) Arrangement of Noctua Active noise cancellation fan system.⁶ Figure 3(b). Performance of Noctua Active noise cancellation fan noise system.⁶

Preliminary testing of active cancellation of fan noise

In preliminary experiments by Prof Fuller, one of the TRW 17 fan sources was used to generate sound and another TRW 17 used to actively control that noise, as shown in Figure 4. Figure 5 presents the results of the measured spectra at the error microphone with control off and control on and a sinusoidal noise disturbance of 10Hz. Large attenuation of the order of 16dB of the noise at 10Hz is apparent. Tests were also successfully carried out with a realistic reference sensor. This system was designed for ANC of infra sound in bedrooms. An operating system was ultimately successfully installed in a large house in Boston, MA and is still operating today.

Figure 4.

Active noise cancellation system test set up.

Figure 5.

Active noise cancellation performance at 10Hz for (a) Control off and (b) Control on.

Testing of the “Self Cancellation” ANC System

A TRW 17 source, as described above, was obtained and set up in Prof. Fuller’s laboratory at the NIA. The set-up is shown in Figure 6. A rod was placed close to the fan and used to generate a fundamental fan tone. For this test, the controller reference signal was generated by placing a microphone located near the fan and used in conjunction with a digital feedforward controller^4,5 already in existence in the NIA laboratory. The controller generated a time varying signal coherent with the reference signal that was amplified and used to drive the vibration of the fan blades. The radiated fan tone was minimized at an error microphone placed in the radiated sound field and used to generate a feedback error signal used to adapt the controller.

Figure 6.

Test set up to evaluate self Active noise cancellation of fan noise using a TRW 17 fan source.

A block diagram of the control approach is shown in Figure 2. A separate monitor microphone was used evaluate the global nature of the ANC reduction. Note that this control approach was an adaptive paradigm^4,5 which could automatically adjust to variations in the fan speed or level. Figure 7 shows an example result from the ANC testing measured at the monitor B&K ½ microphone which was located well away from the error microphone. The tone at 26Hz was calculated to be the fan fundamental tone from the fan rpm and number of blades. With the control turned off, the fan level was measured to be 48.4dB at the monitor microphone as shown in Figure 7(a).

Figure 7.

Results of preliminary testing of self Active noise cancellation of fan noise. (a) Control off (b) Control on.

When the ANC controller was turned on, the sound level was reduced by close to 6dB at 26Hz as shown in Figure 7(b). Data was acquired and processed using a B&K system. Other testing with an independent microphone at a number of positions showed that the ANC reduction was global, or occurred throughout the extended area surrounding the fan. In addition, the drive voltage to the fan blades was measured while the ANC system was operating, to be less than 0.5V p-p. The fan rotation was stopped and the fan blades were then driven with a 0.5V p-p sine wave signal at 26Hz. The radiated sound level at the monitor microphone due to the fan vibration alone was then measured and found to be around 30dB or much less than that needed to cancel the fan tone. This proved that the self ANC effect was due to the combination of the vibration of the fan blades and the fan rotation creating the cancelling noise via airflow modulation. Different fan speeds and frequencies were also tested with similar results. It should also be noted that test situation is not ideal due to a number of reasons such as; (1) a non ideal reference signal was used(in contrast to an ideal one taken from the shaft rotation), (2) the attenuations were also likely limited by the noise floor and reverberance of the background noise in the laboratory (3) the non ideal fan used to demonstrate the ANC system. Higher levels of active attenuation would be likely if these issues were resolved. Nevertheless the preliminary testing unequivocally demonstrated the potential of the self ANC approach for reducing fan noise.

Future work

Future work could consider investigating a number of different areas. The use of a reference signal generated by the propeller shaft motion is likely to give a more highly correlated reference signal without control signal feedback and thus higher attenuation of sound over more fan tones.^4,5 The deployment of virtual error sensors^7–9 is very important as it is not practical to have a physical error microphones in the far field for a drone fan noise control system. The active control of broadband fan noise is also possible since fan blade vibration will produce coherent broad band noise. However, delay and thus controller causality needs to be taken into consideration for this latter area. Finally, a practical means of vibration actuation of the fan blades needs to be developed.

Conclusion

The above work has demonstrated the potential for self active noise cancellation of fan noise tones using vibration actuation of the blades around their axis of support. The results show global noise reduction of the first few fan tones of the order of 5 to 10dB. The required displacement and electrical power to achieve global noise reduction has been to shown to be very small relative to the fan motor power requirements. The technical challenge is to develop a lightweight, efficient and practical method of actuating the rotating blades.

Footnotes

Acknowledgements

The author gratefully acknowledges NASA LaRC for financial support of this work as well as helpful technical discussions with Dr. Noah Schiller of that organization. The loaning of a TRW 17 fan source by Bruce Thigpen of Eminent Technologies is also appreciated. Finally, thanks goes to my student, Arthur Wiedemann for help with formatting and uploading the paper.

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

References

Thomas

Burdisso

Fuller

, et al. Active Control of Fan Noise from a Turbofan Engine. AIAA J 1994; 32(1): 23–30.

Homma

Charpentier

Fuller

. Active-Passive Control of Noise Radiated from Personal Computers. In: Proceedings of INTER-NOISE 2002, Dearborn, MI, August 19-21, 2002.

URL http://www.rotarywoofer.com/

Fuller

Elliott

Nelson

. Active Control of Vibration. London: Academic Press, 1996.

Fuller

Von Flotow

. Active Control of Sound and Vibration. IEEE Controls Magazine, December, 1995.

URL http://noctua.at/en/noctua_anc_project

Fuller

Sauz

Papenfuss

. Active/Passive Control of Portable Generator Set Radiated Noise. In: Proceedings of Acoustics 2012, Freemantle, Aust, 2012.

Bean

Schiller

Fuller

. FEM Based Performance Evaluation of an Active Headrest Using Virtual Sensors in a Diffuse Sound Field, In: Proceedings of Inter Noise, Hong Kong, 2017.

Polen

Fuller

Schiller

. ANC of Radiated Sound from a Small UAS Rotor Using Virtual Error Sensors. NASA Technical Memorandum, Washington, DC: NASA, 2020. NASA TM- 2020-5004166.