 So you can let it running and on the fly you can for example put one to the do filter fair in the live watch and once you keep it running you should you should hear some difference in the south which is coming to the headphones. So what happens if I use the fair filter so do filter fair flag I set up the only thing what happens in the main loop is that this part of the code is executed at line 198 which means I just run these two lines and they are doing the same I'm filtering by fair filter my left and right signal okay and I actually use the left and left and right arrays as a source and destination at the same time so I overwrite the previous signal by the filtered one you tell this to the library when you are initializing the filter so here at the beginning you just need to call once the fair init function where you tell well whether you put a pointer to the handle then you put the how much coefficients you have or how many coefficients you have and then pointer to the coefficients and then pointer to some state array so you can try with the filter and the IR filter as well actually the the cutoff frequency shall be the same so you should not see much difference between the filter and IR filter but what you should see as a difference is the CPU load so well if I disable the correlation for example I've got 1.8 percent just for filtering and if I enable the the filter IR takes one percent for me okay so almost 50% of the fair filter with roughly the same characteristics so usually the IR filters are more performant but you can see here what we do to implement it at the IR filter are usually much more sensitive to rounding the numbers so we have tried to use the Q15 format to implement this filter but the output was always zero because the IR filter are tends to have very large and very small coefficients which are very not nice for the fixed float fixed point algorithms so in this case what we do is to convert the left signal to float then I do the filter and I convert back and still with these conversions and the filtering itself I got less CPU power consumed than with the fixed point maybe you can try later somewhere that if you do the IR filter with fixed point it's very difficult to design a filter which is stable and which works as you would like to so for the IR it's much better to use float specifically for audio applications so again we are on the limit of the fixed point numbers