I use the ADMP441 to detect a sinusoidal in a range of about 2550 Hz in an "audio" signal. At the moment, I drive the mic with a sample frequency of 48000 Hz. The data should be used as an input signal to an IIR and FIR filter. Because I need a 1 Hz resolution, I need to process 48000 samples and that might bring overflow problems. So I want to reduce the sample rate. What might bring the most benefits for the signal quality, using a lower sample rate or to sample at 48 kHz and decimate the input stream by an integer factor? I think oversampling should lower the amplitude of quantisation noise because of spreading over a wider frequency range, right?
A second question. The ADMP441 uses a sigma delta adc. As far as I know, the bit solution depends on the sample frequency and the and the digital filter after the sigma delta modulator. How does the integrated sigma delta adc in the ADMP441? What is the factor of oversampling to generate a constant 24 bit wide data convertion? The data sheet provides no information about that.
Thanks for your answers.