ADAU1761 ADC HP Filter taking too long to settle.

Hi ag77,

I just checked with the design engineer who created that digital HPF in the decimation filter, and he says the corner frequency is ~3 Hz and the time constant should be about 53 ms. I don't know why you're seeing a thirty minute settling time. What is the bias on the microphone?

Where are you measuring the level of the microphone? Inside the DSP core?

Using a DC blocking filter in SigmaStudio should solve the problem, but the digital HPF should make that unnecessary.

Hi Brett,

I am measuring the bias by collecting 16384 audio samples from the ADAU1761 over the I2S bus and taking the average. Using a signed 16 bit word to hold the sample, I find that the average is around 10200 on power up. I repeat this measurement every minute and I find the bias drops from 10200 to 3166 after 34 minutes. In the first 6 minutes, the bias drops from 10200 to 5088 and after that, the decrease in bias appears to be linear.

I did make use of the DC blocking filter in SigmaStudio but it only reduced the bias. What I would like to know is if I am missing a register setting that affects the digital HPF or anything that might be causing this. I have attached the SigmaStudio project file I am using if it helps.

Just another piece of information:

I am also using an externally connected analog microphone. The system can be configured to use either the analog microphone or the MEMS microphone. With the file I attached to the previous post minus the HPF setting and the DC blocking filter, I measured an almost zero bias on the analog microphone while getting the 10200 value for the bias for the MEMS microphone. With the HPF enabled, I was getting a bias of -9800 on the analog microphone although it started out as a much smaller value and just continued to grow in the negative direction.

This behavior is similar to the MEMS microphone except the bias starts out as a small negative value which gradually decreased in value for the analog microphone and for the MEMS microphone, the bias starts out as a large positive value that gradually shrinks. In both cases, the value is decreasing and the magnitude of the change in bias is comparable.

I just tried this out on my board, and unfortunately I cannot duplicate your results. When I power on the board and download the program, I certainly am seeing a large DC offset (about 0.06 or 6% of full scale), but the moment I enable the HPF, it goes away completely. I have tried this with both analog microphones and digital PDM microphones, and the results are always the same. Are you using the EVAL-ADMP421Z with the EVAL-ADAU1761Z? That's my setup.

It seems to work fine using the evaluation boards. Are you using your own prototype hardware?

Hi Brett,

Yes, I am using our own prototype to debug this issue. I did confirm your results this morning with the evaluation boards mentioned so it would seem this is an issue on our side. I will post more on this as I find out in case anyone else runs into this type of issue.

If you wouldn't mind sharing your hardware schematic with us, we could look at your design and see if there's anything that might be contributing to this dc bias issue.