[FEATURE] A possible use case for loudspeaker distortion reduction

[artiko88]

Hey!

I was thinking whether it is possible to capture a model of an active subwoofer, derive an error signal by subtracting an original signal from the one processed by neural-amp-modeler-plugin and adding an inverted version of error signal to the signal chain so that hopefully distortions added by a subwoofer are eliminated by an inverted version of error signal.

If it is a possible use case for neural-amp-modeler, I understand that an interface modification won't be added to the GUI since this is not the intended application of the software. I was just wondering if NAM is capable of training a model that captures non-linearities of an active speaker like a subwoofer and be used for correction. If it is possible, should one modify an input signal and would it be modified?

I know that attempts to model non-linear properties of a loudspeaker have been made and firms like Klippel even developed solutions that they use, so maybe NAM can be a free alternative.

Thanks and sorry if this question is not appropriate.

[sdatkinson]

Fun idea 🙂

I don't think that changes should be made to this repo for it, but the ingredients to try what you described are all here (modulo some data collection work). Please feel free to try it out and let me know what you find!

Cheers!