The conventional approach to acoustic noise control uses passive techniques to attenuate the undesired noise. The passive silencers are effective over a broad frequency range, but ineffective at low frequencies. In case of low frequencies usually active noise control systems are applied rather than passive silencers. The active noise control systems are based on the principle of superposition. Using both active and passive noise reduction technique achieves more efficient attenuation over the complete frequency range. Active noise control (ANC) applications are frequently used to reduce the noise of heating, ventilating, and air conditioning (HVAC) systems. These systems should be modeled with noise reduction in duct. Most common ANC systems have a primary noise generator, a cancelling loudspeaker, a reference sensor and an error microphone. If the noise source is a fan, then beside the periodic noise caused by rotation, the turbulence of the airflow and chafing cause stochastic noises.
I have made an ANC system in which the noise reduction has been achieved in duct, and I measured the efficiency of noise cancellation by algorithms. The turbulence corrupts the coherence between the reference and error signal, so that is why protected microphones have to be used. Better performance might be achieved in noise reduction, if broadband and narrowband noise control structures are used at the same time.
In my Master Thesis I have reviewed the literature references that I need for my ANC project, and then I have written about design and realization of an HVAC system modeled by duct. I implemented an FXLMS and a resonator based ANC algorithm with ADSP-21364 EZ-KIT Lite. The ADSP-21364 board does not have enough inputs, thus I used an eight channel DSP system to run both algorithms at the same time. Thereafter I measured the effectiveness of ANC system using a fan as noise source. In case of a loudspeaker simulated fan noise using both algorithms together means better noise reduction, but in case of real fan noise no improvement had happened. It means I had not experienced any significant difference in using the algorithms together or separately.