ACTIVE NOISE CONTROL IN HEATING, VENTILATION AND AIR CONDITIONING SYSTEMS. Alessandro Cocchi, Massimo Garai & Paolo Guidorzi

Transcription

1 Page number: 1 ACTIVE NOISE CONTROL IN HEATING, VENTILATION AND AIR CONDITIONING SYSTEMS Alessandro Cocchi, Massimo Garai & Paolo Guidorzi University of Bologna, DIENCA Viale Risorgimento, Bologna, Italy alessandro.cocchi@mail.ing.unibo.it Abstract Noise emitted by heating, ventilation and air conditioning (HVAC) systems is one of the main causes of disturbance inside buildings. High frequency noise can be reduced using passive devices, but noise components at frequencies below Hz are hardly reduced in such way. Active noise control (ANC) has its maximum efficiency at low frequency and then it is the ideal complement to passive techniques for achieving acoustic comfort over the whole audible spectrum. In this paper the maximum performance of an experimental active noise control system is investigated. A real scale model of an air conditioning system was built and the ANC system was applied to it. The performance of the ANC system has been tested in a realistic case (HVAC system terminating into a room) and by placing an anechoic termination at the end of the duct. INTRODUCTION For this study, a real scale model of a heating, ventilation and air conditioning system has been built. In the first phase, the model was used in a realistic configuration, with the duct terminating inside a control room (figure 1). In the second phase, the active control system was tested coupled to an anechoic termination (figure 2), in order to obtain the maximum noise cancellation performance.

2 Page number: 2 Figure 1 Control room configuration Figure 2 Anechoic configuration ACTIVE NOISE CONTROL SYSTEM The active noise control system used in this work is based on a commercially available DSP board (EZ-ANC). Two microphones (for sampling the reference and error signals) and a wide-band loudspeaker transducer were used. The feedforward control type was chosen (figure 3).

3 Page number: 3 Figure 3 System configuration In order to test the system performance with the greater variety of noise signals, the usual noise source, i.e. the air fan, was substituted by an artificial noise source: the noise to be controlled was emitted by a loudspeaker connected with a power amplifier fed by a normal PC sound card (figure 3). In this way the ANC system could be investigated when submitted to a set of pure tones combined in many different ways. Additional tests using together the air fan and the artificial noise source were also performed, without significant loss of performance of the noise control system. Figure 4 shows the control algorithm. P(ω) is the primary disturbance component, E(ω) is the error signal, R(ω) is the signal at the reference sensor, A(s) is the primary path transfer function, Y(ω) is the signal generated by the control system, S(ω) is the error signal component due to Y(ω), C(s) is the cancellation path transfer function, Nr(ω) is the measurement noise at the reference sensor, Ne(ω) is the measurement noise at the

4 Page number: 4 error sensor, X(ω) is the control filter input, B(s) is the transfer function of the control signal inverse path. The pseudo-controller transfer function is an IIR filter: Y( ω ) Gs () Hs () = = U ( ω ) 1 GsBs () () (1) Figure 4 Noise cancellation algorithm The system control algorithm attempts to minimize the error signal E(ω). The cancellation path transfer function C(s) is estimated by the DSP board emitting a pseudorandom noise inside the HVAC duct through the control transducer (loudspeaker) and cross-correlating this emitted signal with the signal received at the error sensor; it is represented by a FIR filter. The overall control system is a IIR or FIR filter. The solution is found by a gradient descent algorithm. PHASE ONE: REALISTIC TEST In this experimental phase, the active noise control system was tested inside the HVAC duct configured as shown in figure 1, simulating a real case. The performance of

5 Page number: 5 the control system was evaluated by measuring the residual sound pressure level at the error sensor and at a second microphone placed inside the control room. It was found that the most meaningful parameters influencing the final result are: the probes positions inside the duct, the number of taps of the IIR control filter and the number of taps of the cancellation path transfer function. Some other variables, like the leakage and convergence coefficients of the gradient descent algorithm, are very significant too for obtaining a stable state of the system. Specifically, in this configuration, the parameters values of the control system (for pure tones cancellation) are the following: - sampling frequency: 5.21 khz - control filter number of taps: 15 (forward), 4 (backward) - adaptive algorithm convergence coefficient: adaptive algorithm leakage coefficient: 8 - cancellation path transfer function number of taps: 40 - cancellation path transfer function convergence coefficient: cancellation path transfer function leakage coefficient: 8 The optimal position of the reference and error sensors has been found by maximizing the coherence function between the two microphones. Figure 5 - Coherence function between the two microphones Figure 5 shows the coherence function plot when the distance xr xe (see figure 3) is 3.9 meters. This was found to be the optimal distance for the first configuration. The coherence function gives also essential information about the low and high frequency limits of the system, due both to physical reasons and to the computational power of the DSP board.

6 Page number: 6 Table 1 shows the attenuation obtained for pure tone disturbances. These values are measured with the distance xr xe fixed to 3.0 meters, as this distance gives good results for both the room and anechoic configurations. Frequency (Hz) Attenuation (db) Table 1 Pure tone attenuation due to active noise control PHASE TWO: TEST WITH THE ANECHOIC TERMINATION In a second phase, the active noise control system was tested when the duct ended with an anechoic termination. This test was performed with the purpose of evaluating the maximum performance of the active noise control system. In this configuration, in fact, the sound pressure level of the (apparent) second noise source due to the reflection of the acoustical waves at the end of the duct is drastically reduced, allowing to the control system to work on a single noise source. Figure 6 Anechoic termination

7 Page number: 7 The anechoic termination, shown in figure 6, has a gradual variation of acoustic impedance due to the cut on the top side of the duct and the two wedges of poliurethanic foam. This anechoic termination has a good performance from the 100 Hz to the 400 Hz one-third octave bands. The setting of the control system parameters is similar to that of first case. Only small changes has been operated for adapting the algorithm to the different configuration. Table 2 presents the pure tone attenuation obtained by using the active control system. It can be seen that with the anechoic termination there is a dramatic improvement of the resulting attenuation, proving that in this configuration the active noise control system works at its best. The improvement is particularly remarkable at low frequency. Frequency (Hz) Attenuation (db) Table 2 Attenuation due to active noise control The synergic use of both active and passive systems could be the optimal solution for obtaining the maximum acoustical performance. Table 3 shows the differences between the attenuation obtained using the anechoic termination and the attenuation obtained terminating the HVAC duct in the control room. Frequency (Hz) Attenuation (db) Table 3 Difference of attenuation using the anechoic termination and ending the HVAC duct in the control room

8 Page number: 8 CONCLUSIONS The tests performed using the anechoic termination showed the great potential of active noise control in HVAC ducts. The type of anechoic device used here has small influence on the air flow inside the duct and therefore could be used in air conditioning system where an active control system is present. The performance of the active control has also been measured in the control room (first configuration), using as test signal a low-pass filtered white noise: the overall attenuation was better than 6 db. This is quite normal because of the limited computational power of the DSP board. Future improvement of the research will include the use of a compact anechoic termination in concomitance with an HVAC system in its standard configuration and the optimization of such noise control system. REFERENCES W. B. Conover, Fighting noise with noise Noise Control, 2, (1956) P. Guidorzi, M. Garai and A. Cocchi, Un prototipo di controllo attivo del rumore in condotti a sezione rettangolare, Atti XXVII Convegno Nazionale AIA, Genova, Italia, maggio, (1999) S. Holgersson, Development of an anechoic termination for fan noise measurements, ASHRAE Trans., 74, Part I, Paper No (1974) K. Kido, Reduction of noise by use of additional sound sources, Procedings of Inter-Noise 75, Sendai, Japan, (1975) J. Laumonier and L. Hardouin, An active anechoic termination for low frequencies with mean flow, Acta Acustica, 83, (1997) P. A. Nelson, S. J. Elliott, Active Control of Sound. (Academic Press, London, 1992) S. Snyder and G. Vokalek, EZ-ANC user s guide. (University of Adelaide, Adelaide, 1994)