the conclusion I have reached is to use normal minimum-phase filters and leave time/phase alone, unless using room correction software that supports it (like Dirac)
I have experimented a lot during the past months with all options. My findings (which, if I am not mistaken, corroborate Magnus’s findings and advice):
-
generating “automated” filters from REW measurements did not work for me. In the beginning I ended up with a lot of distortion and at best I was unhappy with the sound quality
-
I had more luck with using REW measurements to create my own few equalisation filters in DSP. An additional advantage is that in this case the CPU taxing is lower and I can still use native DSD (if I use convolution my Roon Server cannot manage the #of taps and I have to convert to PCM)
-
I had a very positive experience with HAF and Thierry. The filters I got with cross talk optimization are eye-opening. But still not a panacea as I find that for some recordings they induce something akin to a partial phase inversion (an elongation of soundstage which with recordings that are anyway using out of phase techniques sound a bit weird). I alternate between “no DSP” and Thierry’s filters, and I have a few of them (with and without an extra bass correction, with and without cross talk optimization)
-
I have also taken the parallel ‘bold’ step of getting a miniDSP SH Studio which (even without room correction) can act as a pure Roon end point (via Volumio) or USB -> SPDIF converter from my EX RoonServer. I was very impressed with the Dirac filters. However, the “internals” of the miniDSP SH are not on par with my reference signal path ( ethernet to Sonore UR and then USB-> optical module of my MSB DAC). This is really a pity as the room correction capability of Dirac appears to be limited by the hardware. In that respect the option of having Dirac working within Roon would for me be the real way forward, as asked by other people as well)