The Intellectual People Podcast - Mitch Barnett

[jtwrace]

Mitch Barnett of Accurate Sound tells us about his journey to become a leading expert in Digital Sound Processing (DSP). Accurate Sound is a remote service using Audiolense, Acourate and Room Eq Wizard (REW).

 

 

 

[ecwl]

Thank you for a fantastic podcast.

 

I have to admit, my experience with “Filter Insertion Loss” is a little different than @mitchco

Yes, obviously with all convolution filters, because of the filter insertion loss, I have to adjust the volume to ensure that I’m comparing one filter vs another at the same volume level.

 

However, I think Rob Watts, the designer of Chord DACs, has mentioned before that from his experience, small signal accuracy both in the digital and analog domain, even in scientifically presumed inaudible SNR levels, can affect depth perception of the music, both in terms of soundstage depth of the music and the 3-dimensionality of the depth of the instruments being played.

 

I have generated my optimized convolution filter in a similar manner where one is at -3dB with less aggressive frequency correction and another is at -7dB with more aggressive frequency correction, leading to a 1% improvement in inter aural coherence coefficient (IACC) and marginally flatter frequency response. Even after compensating for the “filter insertion loss” of 4dB digitally, I can still hear the loss of soundstage depth and instrumental/vocal volume in depth with the -7dB filter. On the other hand, the -7dB filter with the slightly better IACC does lead to better coherence of the instruments and sometimes even marginally better transient attack. I would say for half my music, I prefer one filter and the other half, I prefer the other. 

 

To me, there is no free lunch even with the most advanced convolution filter. I have to admit I don’t have active digital crossovers and it is possible that the effect I’m hearing is not audible in other systems or not audible by other people.

 

At the end of the day, these advanced convolution filters, when properly generated by Audiolense or Acourate by experts such as @mitchco, are simply miles ahead of the alternatives out there. So in many ways, I’m nitpicking here.