ADAU version that supports noise reduction
Thread Summary
The user seeks to convert a sealed ABS plastic enclosure into a sensitive microphone, using a piezoelectric disc and a three-stage charge-coupled amplifier. The final answer suggests using the ADAU1452/ADAU1466/ADAU1467 DSPs for noise reduction, which have 4 serial I/O ports for I2S/TDM interfacing, or the ADAU1761 with on-chip ADC/DAC and a noise removal algorithm. The user clarified the need to subtract structural noise from the speech signal using two sensors.AI Generated Content
Hello everyone. I have a sealed plastic enclosure, 80 x 40 x 20 mm, with a 1.5 mm wall thickness, made of ABS plastic. I need to convert the front panel into a highly sensitive microphone. I used a piezoelectric disc, glued it to the inside surface, and assembled a three-stage charge-coupled amplifier. The sensitivity was decent, but any touching of the enclosure, light impact, etc., immediately drowns out the speech. I've come up with the idea of using a second sensor to remove structural noise and long-range subtraction of one signal from the other, leaving only speech. Which ADAUxxx model can I use for this? Technically, I need two analog inputs and an I2S-type digital bus to connect to the MCU. Thanks in advance for your reply.
Audio Signal Processors
hardware
ADAUxxx
SigmaDSP Audio Processors
Hello Artur82,
You can have a look at ADAU1452/ADAU1466/ADAU1467 DSPs, it has Noise reduction algorithm available for licensing and trial version is there for 30 minutes. This algorithm runs at frequency domain. Here is the link for the Algorithm's info. It has 4 serial input and 4 serial output ports for I2S/TDM interfacing. These DSPs don't have on-chip converters (ADC/DAC), so you will have to use them externally.
Then ADAU1761 is another DSP with on-chip converters (a stereo ADC and DAC), it has a noise removal algorithm, Here is the link for the algorithm which runs on sample based. But this DSP has minimal DSP resource, you will have maximum of 1024 instructions per sample period at 48KHz. Whereas the ADAU14xx DSPs have 6144 instructions per sample period at 48KHz.
You can have a look at these algorithms and other parameters of these DSPs and decide which one is apt for your application. We will be able to assist you further if you have any more questions.
Regards,
Harish
Hi Harish, thanks for your answer.
Perhaps I phrased the question incorrectly. I'd like to subtract the signal from the first sensor (speech + structure-borne noise) from the signal from the second sensor (structure-borne noise).
Please log in to reply.