Adaptive MFxLMS block not training, secondary filter coefficients staying at 0

Dear SigmaStudioPlus Support Team,

We have carefully reviewed your wiki documentation and tested the provided demo program. Despite following the recommended training procedures for both secondary path and LMS coefficients, we are encountering two significant issues:

Secondary Path Training Problem: (we muted input while training)

The training consistently returns values below 1e-6

These values continue to converge toward zero over time

LMS Coefficient Issue: (we redownloaded after trained 2nd path, unmuted)
The coefficients remain static during operation

Even when changing the error microphone position relative to the anti-noise speaker, no coefficient updates are observed

We tried filter length 8 and 32, step size 0.1 and 1e-3. Results didn't show much difference.

Our current setup configuration:

Reference microphone: ~10cm from noise speaker

Error microphone: ~20cm from anti-noise speaker

SigmaStudio diagram implementation follows standard configuration

Below is an image of our setup and sigma studio diagram, with the ref mic ~ 10cm to the noise speaker and error mic ~20cm to the anti-noise speaker. For the current progress, we are just trying to make one linear path work but will later implement three other speaker-mic combinations.

                                Current Setup

An Abstract Top View of the System:

The Sigma Studio Setup:

Secondary Path Values (Tiny)

We would greatly appreciate your guidance on resolving these issues. Please let us know if you need any additional information about our setup or if you would like us to perform specific diagnostic tests.

Thank you for your assistance.

Audio Signal Processors

adau1452

anc

software

Sigma Studio+ 2.0.0

SigmaStudio+ 2.0

SigmaDSP Audio Processors

MFXLMS

Show More