This paper presents a study overview of the Active Noise Cancellation (ANC) technology and demonstrates the technology with a real time setup. The paper highlights the innovation and challenges in demonstrating the technology. In the process the core Adaptive signal processing algorithm is explained in detail.
Active Noise Control Real Time Demonstration Srikanth Konjeti Srikanth.Konjeti@harman.com
Abstract: This paper presents a study overview of the Active Noise Cancellation Noise Error (ANC) technology and demonstrates the Microphone Microphone technology with a real time setup. The paper highlights the innovation and challenges in Noise demonstrating the technology. In the process Source the core Adaptive signal processing algorithm Anti noise Loudspeaker is explained in detail. ANC Controller Keywords: Active Noise Control (ANC), Filtered –XLeast Mean Squares (FXLMS), Real timeExperiments, Secondary path Estimation “Fig.1. Active Noise Control Setup” 1. IntroductionActive Noise Control (ANC) is a technology that is used 2.1 Filtered-X LMSin controlling the noise in the real life scenarios. The main e(n) x(n) d(n)objective is to generate anti noise which is out of phase P(z) +and equal in amplitude to the noise under test. Thetechnology uses the adaptive filter and adapts to the y(n) y’(n)changes in the noise characteristics and generates antinoise using the feed forward control mechanism. w(z) s(z)ANC achieves noise reduction particularly at lowfrequencies. The applications include Automotive, x’(n)Appliances, Industrial and Transportation ^ LMS s(z) 2. Overview “Fig.2. Block Diagram of FXLMS”Acoustic noise and the noise related problems have been The block diagram of the FXLMS is shown in Fig.2 Theon the rise since the industrial revolution and the advent of primary noise x(n) passes through the primary path P(z)machinery. The usual way to deal with noise reduction is and reaches the error microphone. The captured primaryto stuff the construction with bulky material which is noise x(n) is filtered with the adaptive filter and y(n) isusually costly and is ineffective in reducing the low generated. The anti noise y(n) will be changed because offrequencies. the secondary path between the loudspeaker and the error The Active Noise control techniques deals with the low microphone. To compensate for the effects of thefrequencies effectively. The technique is primarily based secondary path the transfer function of the secondary pathon the superposition of the noise with the noise of equal is measured and placed in the path of the LMS algorithm.amplitude and opposite phase resulting in a null at thepoint of cancellation The accumulated input The overview of the ANC technique is represented in 𝑿 𝒏 = [𝒙 𝒏 , 𝒙 𝒏 − 𝟏 , 𝒙 𝒏 − 𝟐 … 𝒙(𝒏 − 𝑳 − 𝟏)] 𝑻the Fig.1. The Primary Noise is captured using a The accumulated Outputmicrophone or sensors and the anti noise is generated 𝒀 𝒏 = [𝒚 𝒏 , 𝒚 𝒏 − 𝟏 , 𝒚 𝒏 − 𝟐 … 𝒚(𝒏 − 𝑳 − 𝟏)] 𝑻using the Anti Noise Loudspeaker. The resultant noise And the filter tapsafter cancellation is captured by another microphone 𝑾 𝒏 = [𝒘 𝒏 , 𝒘 𝒏 − 𝟏 , 𝒘 𝒏 − 𝟐 … 𝒘(𝒏 − 𝑳 − 𝟏)] 𝑻called the Error microphone (Fig.1). The error The secondary pathmicrophone acts as the feedback mechanism for the ANC 𝑺 𝒏 = [𝒔 𝒏 , 𝒔 𝒏 − 𝟏 , 𝒔 𝒏 − 𝟐 … 𝒔(𝒏 − 𝑳 − 𝟏)] 𝑻controller. The filtered primary noise through the secondary path The noise varies its amplitude, frequency with time and 𝑿′ 𝒏 = [𝒙′ 𝒏 , 𝒙′ 𝒏 − 𝟏 , 𝒙′ 𝒏 − 𝟐 … 𝒙′(𝒏 − 𝑳 − 𝟏)] 𝑻the ANC keeps track of these changes and generates antinoise using the adaptive filtering techniques. The error between the primary noise and the anti noise The anti noise loudspeaker is present in the path that the with the secondary pathnoise takes to reach the error microphone calling it a Feed 𝒆 𝒏 = 𝒅 𝒏 − 𝑺 𝑻 𝒏 ∗ 𝒀(𝒏)forward cancelling technique. The Null zone is created at The anti noise is generated from the adaptive filterthe error microphone. The primary noise and the noise captured by the error 𝒚 𝒏 = 𝑾 𝑻 𝒏 𝑿(𝒏)microphone is fed into the LMS based adaptive filter The coefficients of the adaptive filter are continuouslywhich varies its filter coefficients to minimize the mean adapted as the followingsquare error between the primary and the anti noise. 𝑾 𝒏 + 𝟏 = 𝑾 𝒏 + µ 𝒆 𝒏 𝑿(𝒏) 𝑿′ 𝒏 = 𝑺 𝒏 ∗ 𝑿 𝒏
2.2 Real Time Experiment Setup An impulse response is used in measuring the secondaryAn experimental setup to demonstrate ANC is shown in path response. Fig.5, 6 shows the responses and noise atthe Fig.3. A pair of Harman Kardon HKTS speakers are low frequencies. This method of measurement in the noisyused in the experiment. The speakers are connected to an environment is inefficient and destabilizes the LMS filter.amplifier and the noise is played from the computer. Onespeaker is used as the source of noise and the secondspeaker is used to generate anti noise. A Behringermicrophone connected to the audio card is used as theerror microphone. The audio card is connected to the PCvia the USB. The DSP software runs on the PC as a VSTplug-in. The noise is sent through the PC to a loudspeaker andalso to the VST plug-in. The signal captured by the errormicrophone is also fed to the VST plug-in software. TheDSP software on the PC analyzes the noise source, the “Fig.5. Impulse Response of the Secondary Path”error signal and generates the anti noise that is fed to thesecond loudspeaker. “Fig.6. Frequency Response of the Secondary Path” To overcome this problem we used sine waves with the cancelling frequencies of interest as the source to measure the transfer function. This is robust to the external noises and accurately measures the transfer function at the frequencies of interest. Fig.7 shows the secondary path response measured with a sine wave of 200Hz. “Fig.3. Error Microphone Setup”Challenges 1. Measuring the Secondary path Response 2. Stability of the LMS algorithm “Fig.7. Frequency Response of the 2.3 Secondary Path Transfer Function Secondary Path with 200Hz Sine wave”The path from the Anti noise loudspeaker to the ErrorMicrophone is called the Secondary path.. As Shown in 2.4 Resultsthe Fig.4, white noise is played through the speaker and It is common in the industry and automobiles to findthe signal is captured through the error microphone. Both steady noise. It has audible discrete frequencies and ofthe signals are fed to the LMS algorithm and over time the steady amplitude. So the sinusoids are used as noise herefilter converges to the transfer function of the secondary are played through the computer and connected to thepath. amplifier and to one of the speaker. The anti noise is This method of measuring the transfer function is played from the computer to the loudspeaker. A switch iseffective when the measurement is taken over a very quiet placed in the VST plug-in (ANC OFF/ON) of the DSPenvironment or the low frequency external noise will software to turn the algorithm OFF/ON. The Fig.8, 10derail the response shows three plots X(n) Anti Noise Error a) Sine wave as the noise input played through the Speaker Microphone speaker. White peaker Secondary Path b) The Anti noise generated by the VST DSP software Noise / Impulse and played through the anti noise speaker y(n) c) The Error signal captured by the error microphone The gaps in the anti noise plot shows the periods of LMS ANC ON and OFF. It clearly establishes that when the switch is OFF the error signal increases and when the “Fig.4. Measure Secondary Path Transfer Function” switch is ON the error signal decreases.
The anti noise plot shows an initial period of 3-4seconds where the adaptive filter ramps up (Adaptationstage) to start cancelling the noise. In this period there isno change in the error signal. Once the filer adapts to thenoise it remains steady and generates anti noise. The convergence parameter of the LMS filter plays animportant role in the stability and performance of thealgorithm. The parameter is empirically tuned to have anoptimum performance and maintain stability.The empirical value is shown in Table.1 “Fig.10. Multi tone Sine wave Input, Anti Noise, Error Signal” “Fig.8. 200Hz Sine wave Input, Anti Noise, Error Signal” The frequency response in Fig.9 shows the sine wavewhen the ANC is OFF and ON. There is a reduction of~50dB of the sine wave when the ANC is ON “Fig.11. Frequency Response of the Multi tone Sine wave with ANC OFF/ON” The frequency plot shows a significant reduction in the three frequencies when the ANC is switched ON. 3. Conclusion The paper presents an insight into the ANC technology and demonstrates the potential using a real time setup. The technology has immense application in the automobile industry and currently adopted in the automobiles to suppress Engine noise and Road Noise. The real time setup can be expanded to a real life application to create “Fig.9. Frequency Response of Sine wave with ANC OFF/ON” silent zones around the head of a person in the office and Input ANC OFF ANC ON Reduction external environments. Amplitude dB Single Tone 20 -30 50 4. References Multi Tone 10, 10, 10 -23, -10, -13 33, 20, 23 . SEN M. KUO AND DENNIS R. MORGAN, “Active Convergence 0.0002 µ Noise Control: A Tutorial Review”. “Table.1. Performance of ANC” . Lichuan Liu, Sen M. Kuo, and Kishan P. Raghuathan. “Active Noise Control for Motorcycle Helmet”. Multiple SinusoidsThe ANC experiment is carried on multiple sine waves of150+200+250 Hz which sounds like the motorbike noiseon the road.