Dear developers,
We use the ICS-43432 in an embedded audio application with a Blackfin BF527 from Analog Devices. We use the I2S interface to read data from the mems. We observe unusual errors with the device.
Startup We have multiple different startup patterns in the data. The specified behavior of the microphone is that after 2000 samples the digital filter is operational and that the data is actual recorded audio data. We observe the following when we apply power and fs to the device.
- We have a normalised signal after 2000 samples (as it should be)
- We have a long swirling curve (with blocky signal) which sometimes takes >12000 to normalise
- We have a sharp drop to a low value (sometimes twice) and then it normalises slowly
We know that the digital filter takes time to be setup but this behavior is unspecified and unpredictable. It happens at random. The strange part is that sometimes two or more data channels overlap completely (two different mems creating and sending exactly the same sample). We have tested and measured every signal and are stumped that it doesn't work.
Blocky Patterns Next to the startup problems we have strange behavior in the data. We have blocky patterns. The mems will output the same value for some time. Then it will change to a different value and do the same. We would suspect this in a parallel connection but it is very strange with the serial connection. The blocky patterns also occur randomly accross all of our 8 microphones. It only disappears when the device is unplugged and plugged back in (completely removing power). The blocky pattern is also visible with the long curvy startup behavior. It can happen with 2, 4 or 8 channels at the same time. We would like some explanation to guide us in the right direction to solving this problem.
To view the plotted data please rever to the Google Drive folder here: Drive. We have also send an email to the developers support address. It is in de folder and contains more information.
Our dataclock is at 3.072 MHz and our FS clock is at 48kHz. Our sample size is 24 bits and we take the first 16 bits of the two's-complement sample. We use I2S as specified in the datasheet and we have verified that we adhere to the datasheet. We have a previous design which has no unspecified behavior, this design incorporated the INMP441 Digital Mems. The connections are inspired on this application notice: High-Performance-Digital-MEMS-Microphone-Standard
We have a deadline begin February and we need these errors sorted. Please feel free to ask further questions or give any advice.