One-third octave spectrum calculation to meet the attenuation limits of a Class 1 analyzer (IEC 61260-1:2014) for low-power noise sensors

Thanks to the advances in technology and the cost reductions of electronic components, nowadays it is possible to produce low-cost and low-power acoustic sensors with extended capabilities such as the one-third octave band spectrum calculation. However, there are some doubts about the accuracy of these devices. Thus, to meet the attenuation limits of a Class 1 analyzer for low frequencies as stated in IEC 61260-1:2014 normative, this research proposes two algorithms to calculate the one-third octave band spectrum, both algorithms are based on the Power Spectral Density (PSD). The first one consists of a subdivision of frequency bins, where a PSD bin is divided into smaller parts, gain adjusted, and assigned to its corresponding one third-octave band. The other one is based on signal decimation-accumulation, where samples, that are processed every 125 ms due to the sensor design, are first decimated and then accumulated to increase the frequency bin resolution. Both algorithms are programmed into the sensor and the power consumption and acoustical performance are evaluated and compared to a typical multi-rate filter bank. It is observed that both proposed algorithms improve the performance of the low-cost sensor in terms of energy consumption and accuracy.