For the last decade or so I've experimented with high frequency dips modeled after the Linkwitz Gundry dip (generally characterized as 3dB down at 3kHz with Q of 2, which is between a half octave and a full octave bandwidth) but on the basis that the issue is excess high frequency reverberation in playback generally, and not just in the region of cochlear resonance centered at 3kHz. This reverberant field causes more high frequencies to enter the ear canal than would happen in a live performance with the music sources in front.
I've generally started a roll off around 2kHz and going up to 12kHz or so, above which some added highs are beneficial. But I also came to feel it imparted a certain dullness. So I decided to rethink the "Peterson Dip."
Now I'm trying a dip centered at 4kHz (more like in BBC speakers, btw) with an bandwidth of 2 octaves, and limited to 3dB maximum. The contour points are then 2kHz and 8kHz, which includes what I consider most important--the "metallic" range around 6kHz, without necessarily rolling back the higher frequencies which add "air" or suppressing 1kHz which would remove clarity. The Q is a comfortably low 0.667 with hardly any overshoot whereas a Q of 2 has significant ringing. Previously I used 3 filters with a octave bandwidths or less to achieve the 2-12kHz dip, but unless you are truly cancelling resonances, which this is not, I think fewer filters are better.
Noting that my Acoustats measure fairly flat in the range 1kHz-16kHz, I decided to simply apply the new dip on top of their pre-existing response without getting deep into achieving the desired response acoustically for the time being.
It seems to provide about the desired effect on taming excess brightness and making it easy to listen at a higher levels too.
