Real-time ANC Simulink Implementation of Measure Impulse Responses

I am attempting to implement the model based implementation of the ANC system as described in https://www.mathworks.com/help/audio/ug/active-noise-control-with-simulink.html . More specifically, I am attempting the following system:

In the description of the system, it is stated the the primary and secondary path FIR filters are chosen arbitrarily and can be updated with actual impulse responses of a real environment to make the simulated system more realistic. This is what I am attempting to do. I have measured frequency response measurements for a primary and secondary path of a real world envrionement in REW and exported a text file containing values of h(n) of the respective windowed impulse responses of the two paths. These impulse responses start at t=0. However, they are of length n = ~22,000. I assume that these are too long and are not what is supposed to be implemented for the FIR filters of the paths in simulink. Therefore, my question is how to go about implementing real world impulse responses for the primary and secondary path FIR filters in the simulink model?

[Mathieu NOE]

hello

I see rwo reasons your FIR filters are too long :

• excessive sampling rate - which frequency range are you targeting ?
• the model (real or not) used to create the FIR filters has a too long decaying rate; you can either truncate your FIR length (I would try first with 256 samples for example) , especially if the trailling sections contains mostly zeroes (or noise) , or apply a exponential decaying window to "force" the IR to be shorter

[Jimmy Lapierre]

Mathieu is right. You can also replace delays in the response by a delay block (I almost said initial zeros, but they won't be pure zero in a real-world measurement). You can also increase the length of the LMS update filter. In any case, as a last step you will probably need to tweek the learning rate of the LMS update.