ADAU1452 Acoustic Echo Cancelling Issue

Hi Paul,

We are actively porting the ADAU1442 algorithm to this new processor, the NLMS filter for ADAU1452 will be included in a future update,

Regards,

John

Hello jmccarty,

SigmaStudio's 3.11.1 beta have this NLMS block for ADAU1452 but seems it is not functioning well. Pin0 and Pin1 are inverted, I believe, and there is no TC in in ADAU1452 as compared to the ADAU1761 chip. There's no documentation as far as I can search, so no clue as to how it works, except for some forum topics before related to this.

Using ADAU1452's NLMS block, I have tried BrettG's model design from here: https://ez.analog.com/message/26410#26410

but I got no output (pin0 &pin1 already inverted). ADAU1761's NLMS block works perfectly fine on this design model from BrettG.

Any thoughts as to what goes wrong on this NLMS ADAU1452 block?

Regards,

Dan

Hi Dan,

The ADAU1452 NLMS implementation is not exactly the same as the one on ADAU1761 even though it looks almost identical to the ADAU1761. I apologize as this is obviously very confusing without any documentation. Basically both the input and output pins are swapped between the 2 implementation.

The ADAU1761 block internally subtracts the "unknown signal" from the output of the NLMS filter and this is the "Signal Output" pin. The orange pin is the output of just the adaptive NLMS filter.For the ADAU1452 implementation the main output is the NLMS filter output (just the filtered far-end signal, equivalent to the orange pin of the ADAU1761 verion). The orange pin is the error, or the far-end signal with undesired echo plus the near-end signal.

You weree correct that the input pin functions are swapped from the ADAU1761 version. To create an equivalent flow to the ADAU1761 example, you can either just use the "error" output directly, or you can subtract the output of the FIR filter from the unknown signal which produces the same result.

Just so you are aware, the ADAU1452 filter calculates "instantaneous energy" of the input state filter update, while the ADAU1761 calculates an RMS value (that is the purpose of the time constant tc). The behavior of these blocks is very similar but will not be identical.

Hi John,

Very helpful information about NLMS in ADAU1761. Would it be possible for you to upload the equivalent block diagram of the block for ADAU1761? It would be helpful in order to finish my work with confidence and have the proper documentation.

Thanks!!!

Ariel.

This question has been assumed as answered either offline via email or with a multi-part answer. This question has now been closed out. If you have an inquiry related to this topic please post a new question in the applicable product forum.

Thank you,
EZ Admin