THE DESIGN OF A 4TH-ORDER BAND PASS FILTER USING SALLEN-KEY TOPOLOGY TO REDUCE NOISE IN THE VOICE SIGNAL.

• Main Author: Herianto
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/67980
• Institution: Institut Teknologi Bandung

The interface system has developed rapidly since the computer was first invented, starting from the use of buttons until now the voice interface system has been developed. Using voice to interact with machines provides many advantages such as increased productivity and efficiency. However, the voice interface system also has various limitations. One of them is background noise from the environment which reduces the level of clarity of the spoken sound so that the machine cannot interpret it. In this final project, a band pass filter has been designed to attenuate background noise from the voice signal so that the sound clarity level can be increased. There are two types of filters, namely analog and digital filters. The analog filter is designed using Sallen-Key topology and simulated in Proteus software to obtain the frequency response bode plot. The digital filter is designed using the Fourier transform principle in Matlab software. After obtaining a frequency response bode plot according to the desired specifications, both filters were tested using sound recordings of several potential noise generators in everyday environments. The filter that has been designed has a 4th order filter rolloff, namely 23.53 and 24.61 dB/octave for analog filters and 23.9186 and 23.8781 dB/octave for digital filters. Both filters are capable of reducing noise outside the frequency range of the 300-3500 Hz sound signal. The test results show that the gain of the digital filter is 1.68% smaller than that of the analog filter. This result is caused by the recording sample used contains more low-frequency spectrum. At low frequencies, digital filters are still able to calculate accurately while analog filters are less precise because of the electronic components used. The analog filter design made in Proteus is implemented in hardware form. The resulting analog filter hardware has a fairly flat passband in the frequency range of 1-4 kHz with a maximum ripple of 1.09 dB. The rolloff filter shows a fairly appropriate value of 22 and 24 dB/octave. The filter hardware has not been tested using voice signal recordings. However, the filter specifications obtained are in accordance with the simulation results so that the filter hardware is expected to be able to reduce noise outside the frequency range of the sound signal.