I tested the "LMS Norm" module by filtering a noise signal in matlab with 16 random, real coefficients, then running the input and output signal through the LMS norm module to see if it could determine the 16 coefficient values. It did not work, until I realized the produced coefficients seemed to be limited to the range +/-1. When I reduced my original coefficients to be in this range the LMS Norm module produced accurate estimates of each. If the module is a "floating point" module, why would its coefficient range be limited to +/- 1? There is no indication as such in the online help documentation. Perhaps I am using it incorrectly?
12:34pm
Hello bw,
Let me please email you on the side about this.
Thanks,
Kevin
1:07pm
We have solved this issue. The limitation to +/-1 was due to my converting to fract32 and saving the coefficients via output to .wav file. Thanks to Kevin Kluka of DSPC for helping sort this out.
1:53pm
Is there a way to pause LMS coefficient updates? I'm not seeing how to do this with the available inputs. Actually using SbNLBS.
4:37am
Hello jivey_8997,
Could you please provide a few more details about what you are trying to do? Are you trying to pause the LMS coefficients in real time while continuing to process audio? Or perhaps save the coefficients of a certain state to an array? Or perhaps simply slow the adaptation time?
Thanks,
Kevin